Mosaicing of acoustic camera images

合集下载

声场重构中消除鬼影和提高重构精度的方法

声场重构中消除鬼影和提高重构精度的方法

文章编号:1006-1355(2013)04-0200-07声场重构中消除鬼影和提高重构精度的方法付强1,2,黎敏1,2,樊悦1,2,魏龙1,2(1.北京科技大学新型飞行器技术研究中心,北京100083;2.北京科技大学机械工程学院,北京100083)摘要:在利用波束形成算法进行声场重构时,容易出现鬼影现象,给重构声场的准确辨识带来极大困难。

在对波束形成算法进行理论推导后,获得麦克风阵列的结构参数与鬼影产生的关系,进而获得消除鬼影的方法;并在此基础上,优化麦克风阵列的布局形式,设计阵列的最大孔径比,可以有效提高重构精度。

通过纯音实验和窄带噪声实验,验证了消除鬼影和提高声场重构精度方法的有效性。

关键词:声学;声场重构;波束形成;鬼影;重构精度中图分类号:TB132;TB534+.3文献标识码:A DOI编码:10.3969/j.issn.1006-1335.2013.04.042 Method of Eliminating Ghost Images and Improving Reconstruction Precision in Acoustic-fields ReconstructionFU Qiang,LI Min,FAN Yue,WEI Long(1.Reasearch Center for Aerospace Vehicles Technology,University of Science and Technology Beijing,Beijing100083,China;2.School of Mechanical Engineering,University of Science and Technology Beijing,Beijing100083,China)Abstract:The ghost images can occur easily in the reconstruction of acoustic fields with the use of beamforming algorithm,which leads to a terrible difficulty for accurate identification of the reconstructed acoustic fields.After detailed deduction of the beamforming algorithm,the relation between the shape parameters of microphone array and the ghost-images occurrence was obtained,and the method for eliminating the ghost images were attained.It was found that after the elimination of ghost images,the reconstruction precision would be improved if the value of D/λwas increased by means of optimization design of the array’s configuration.Finally,the effectiveness and feasibility of this method was verified by some experiments of acoustic-field reconstruction of some pure sound and narrow-band noises.Key words:acoustics;reconstruction of acoustic fields;beamforming;ghost image;reconstruction precision收稿日期:2012-08-29;修改日期:2012-12-10项目基金:国家重大科学仪器设备开发专项:(2011YQ14014507);中央高校基本科研业务费专项资金:(FRF-AS-12-003);中央高校基本科研业务费专项资金:(FRF-MP-12-002A)作者简介:付强(1983-),男,山东日照人,博士,目前从事高温声场测量与重构、材料与构件声疲劳研究。

声光超分辨率成像原理

声光超分辨率成像原理

声光超分辨率成像原理Super-resolution imaging is a technique that enhances the resolution of an image beyond the typical limit of a sensor or optical system. In the case of super-resolution acoustic imaging, the technique involves using multiple microphones to capture sound waves from different angles and distances in order to reconstruct a higher resolution image of the source of the sound.超分辨率成像是一种技术,它可以提高图像的分辨率,超越传感器或光学系统的典型极限。

在超分辨率声学成像的情况下,该技术涉及使用多个麦克风从不同角度和距离捕捉声波,以重构声源的更高分辨率图像。

The principle of super-resolution imaging involves taking multiple low-resolution images of a scene or object and then using computational algorithms to combine and enhance the details to create a single, high-resolution image. This can be achieved through various techniques such as interpolation, deconvolution, and deep learning-based super-resolution.超分辨率成像的原理涉及拍摄场景或物体的多个低分辨率图像,然后使用计算算法将这些图像合并并增强细节,以创建一张高分辨率图像。

Boom AURA智能视频会议条说明书

Boom AURA智能视频会议条说明书

FEATURES The Boom AURA features 4K Ultra HD, 10x digital zoom and built-in AI functions that include highly responsive facial detection and sound source localization providing intelligent auto-framing and voice-tracking. Six digital array mics provide 6 meters of pickup range ensuring crisp audio and an optimal intuitive response. The all-in-one AURA delivers a smart experience for your video conference calls. Intelligent hardware makes better meetings. Simply.Boom AURAan intelligent video bar for superior meetingsIntelligent & versatileChoose between a variety of AI features to optimize the meeting experience. Auto-framing seamlessly centers and frames participants while voice-tracking uses sound source localization to frame and track active speakers for a dynamic meeting experience that incorporates all participants with ease.Quality where it counts8MP CMOS image sensor provides a 4K Ultra HD image projecting a professional, sharp image in every meeting.E xcellent audioHigh-fidelity audio with a 48kHz sampling rate provides lossless audio and full spectrum of sound for meeting participants. Acoustic EchoCancelation (AEC) and Automatic Gain Control (AGC) further enhance the experience and work to ensure Full Duplex quality at a distance of 6meters.A lens with a view120° wide angle field of view ensures all participants are seen, even in small meeting spaces.P lug & Play. Really.Mac or PC? Zoom, Teams, Webex, or Meet? With USB 3.0, simply connect and enjoy your favorite video conferencing platform.Contact us:CameraImage sensor : 4K 8MP CMOS image sensor Lens: 4K glass Pixels: 8MP, 16:9View angle: 120° diagonal, 107°height, 74° vertical Zoom: 10x digitalFocus: Advanced AI auto-focusSNR: ≥55dBResolution: 3840 x 2160 Video features: Brightness, definition,saturation, contrast, white-balance, low light optimization AudioMicrophone : 6 built-in digital array microphones, 20ft/6m rangeAudio processing : AEC, AGC, ANS, sound source localization DNR : 2D & 3D Speaker : 2 x 7WDetailsVideo output: USB 3.0Temperature: 15°F -+120°F(-10°C - +50°C)Dimensions: 23.5in x 3.4in x 3in (598mm x 86mm x 77.5mm)Weight: 4.17lbs/1.9kgMount: wall mount bracket/standOperating system: Windows® 7,Windows 10 Mac OS X® 10.10or higher Hardware requirement:2.4 GHz Intel@ Core 2 Duo processor or higher,2GB memory or higher, USB 2.0port (USB 3.0 for 4K)Cable length: 16.4ft (5m)In the box: AURA video bar, USB 3.0 cable, remote, power cord, wall-mount bracket and quick guideBoom AURA About Boom Collaboration。

Panasonic WV-S4550L 360度防烟摄像头说明书

Panasonic WV-S4550L 360度防烟摄像头说明书

WV-S4550L PJiA H.265 360-degree Vandal Resistant Outdoor Dome CameraPanasonic WV-S4550L captures the highest quality images in even very challenging and dynamic environments. In particular, the image of the person's face or object at the edge of the fisheye is clear with less distortion.Intelligent Auto (iA) monitors scene dynamics and motion to adjust key camera settings automatically in real-time reducing distortion such as motion blur on moving objects. Adopting H.265 Smart Coding technology, bandwidth efficiency is intelligently increased for longer recording and less storage. Out of the box, the camera supports full data encryption streaming and is compliant to FIPS 140-2 Level 1 standards to keep your video secured.Extreme image quality for evidence capturing under challenging conditions-Clear and less distorted image of the person's face and objects at the edge of the fisheye -Instant response to sudden light changes like tunnel entry and exit -Auto Shutter speed control for fast moving objects-Super Dynamic 108dB for backlit situations and shadows on night streets-Built-in IR LED to produce a clear monochrome image in zero lux conditions with 10 m (33 feet) irradiation distance -Environmental durability : EN50155, IP66, IK10, 50J compliant and Dehumidification deviceExtreme H.265 compression with new Smart Coding-Longer recording and less storage space compared to any H.264 based compression techniques-New self-learning ROI* encoding (Auto VIQS) dynamically detects motion areas to keep vehicles and humans in good picture quality while lowering your bandwidth * Region of InterestExtreme Data Security-Full encryption SD card edge recording to keep your data safe -FIPS 140-2 Level 1 compliant-Full end-to-end system encryption with supported VMS and devices to protect from IP snooping/spoofing and detect data alterationComplete with powerful analytics built-in*-Heat map : Visualization of people's traffic pattern and staying times-People Counting : Statistics data on the number of people entering and leaving a specific zone-MOR (Moving Object Remover) : Monitoring of only the surrounding environment by removing people and other moving objects from video *To enable analytics, please add the optional WV-SAE200 software.•5 Megapixel images up to 30 fps • iA (intelligent Auto)• Super Dynamic 108dB • H.265 Smart Coding• IP66, IK10, 50J compliant, Dehumidification deviceKey Features•Retail / Bank • Education / Hospital •Building •Transportation (Airport / Train, Subway station)•LogisticsApplicationswith Base Bracket(Made in JAPAN)DISTRIBUTED BY :/PanasonicNetworkCamera(2A-201DA)• Masses and dimensions are approximate. • Specifications are subject to change without notice.Important– Safety Precaution : Carefully read the Important Information, Installation Guide and operating instructions before using this product.– Panasonic cannot be responsible for the performance of the network and/or other manufacturers' products used on the network.AppearanceUnit : mm (inches)Optional Accessories For WV-S4550L only.i-VMD is possible to detect objects in the specified area by advanced video analysis technology.i-VMD : People Counting, Heat-map,MOR (Moving Object Remover),Intruder Detection, Loitering Detection,Cross line Detection, Object Detection, Scene change DetectionPlug-in Software for i-VMDWV-SAE200Notification sent to the monitoring screen Trademarks and registered trademarks– iPad and iPhone are trademarks of Apple Inc., registered in the U.S. and other countries.– Android is a trademark of Google Inc.– ONVIF and the ONVIF logo are trademarks or registered trademarks of ONVIF Inc.– All other trademarks identified herein are the property of their respective owners.(This bracket requires WV-Q186 or WV-Q124.)(This bracket requiresWV-Q185 or WV-Q122A or (This bracket requires WV-Q185 or WV-Q122A.)(This bracket requiresWV-Q185 or WV-Q122A.)*2 Auto VIQS, i-VMD, can not be used at the same time.*3 Transmission for 2 streams can be individually set.*4 Only use AAC-LC (Advanced Audio Coding - Low Complexity) when recording audio on an SD memory card.CD-ROM for descriptions of how to switch the output. (factory shipment : NTSC monitor)When using the base bracket。

Advanced 3.2 ReleaseNote(English)

Advanced 3.2 ReleaseNote(English)

4). Motic Images Advanced 3.2 needs a separate dongle to run (Motic Images Advanced 3.1 dongle can not run Motic Images Advanced 3.2);
5). Motic Images Advanced 3.2 is compatible with Motic Images Advanced 3.1, i.e., Motic Images Advanced 3.2 dongle can run Motic Images Advanced 3.1.
8. DIS module may not work appropriately under Windows98.
9. When DIS is in use, do not exit during the file transfer period, otherwise it may cause DIS unable to close.
12. To make better effect, if with M1300 device, please select MoticTek option in the Setting window ; if with M1350 device, please select Motic1350 option.
11. If the captured image in the main program has the same size as the captured image in the Capture module, please select Auto for the Image Size option of the Setting window in the main program.

Fishman F1 Aura

Fishman F1 Aura

WelcomeThank you for making Fishman a part of your acoustic experience. We are proud to offer you the fi nest acoustic amplifi cation products available; high-quality professional-grade tools to empower you to sound your very best. We areconfi dent F1 Aura will both enhance and inspire your music making.Despite its simple and clean appearance, F1 Aura features a powerful set of tone shaping controls and programmability. We urge you to read through this user guide and spend some time getting familiar with its operation, so that you may easily realize the system’s full potential.Aura Acoustic Imaging Technology uses digital algorithms developed in Fishman’s audio laboratories to restore a studio-miked sound to an acoustic instrument. To achieve this, we’ve recorded this instrument using world-class microphones and techniques to capture an “Image” of the natural sound thatit emits when miked in a professional studio. This Image, when recorded direct or played through an amp, mixer or PA, blends with the undersaddle pickup to produce an incredibly accurate recreation of the original recording.Play / EditF1 Aura operates in two modes: Play and Edit. Play is for selecting the most frequently used controls while Edit gives you access to many more features.PlayWhen you plug in, F1 Aura powers up into Play mode. In Play mode, Volume, Blend and Phase can be adjusted. The chromatic Tuner and the automatic Anti-Feedback circuit can also be activated.EditPress and release the Edit knob to enter Edit mode; the tuner’s green in-tune LED will light solid. Once in Edit, press the Edit knob repeatedly to step through the parameters. Each parameter is displayed using a single letter to represent its function (see page 15). Adjust its value by turning the Edit knob. A number is displayed and the tuner’s sharp/fl at lights show positive or negative values. Note: F1 Aura is programmable and automatically saves your settings.To exit Edit and return to Play mode, wait 10 seconds for the display to go dark, or press + hold the Edit knob for 2 seconds. You may also immediately exit by simultaneously pressing both the Edit and Volume knobs.5Play Mode Controls (cont.)BlendTurn Edit • Turn the Edit knob without pressing on it and the balance between pickup and Image is adjusted. A setting of P = 100% Pickup signal; 0 = a 50/50 pickup/Image blend; I = 100% Image signal.Suggestions• For live performance try backing off the Image by setting Blend to about 2 or 3 (about 65% pickup).• For recording, try blending in more Image for a realistic acoustic sound. Automatic Anti-FeedbackUse this search-and-destroy Anti-Feedback circuit in addition to Phase to control feedback during a performance. F1 Aura’s automatic Anti-Feedback circuit can apply up to three separate notch fi lters, which are very precise tone controls that reduce only a tiny piece of the audio band. When activated, the fi lter locates and reduces the problematic resonances associated with feedback.8While the Anti-Feedback control is very effective, it’s best if you spend some time while setting up before a performance and catch any issues before you begin to play. With some practice, you’ll fi nd you can also use it to “fi x” any resonant notes that may stand out in a particular venue.Using the automatic Anti-Feedback control:1. Press + hold both Edit and Volume for 2 seconds. The tuner display will fl asha “1” to indicate it is searching for the fi rst feedback.2. Turn up the Volume, then either dampen the strings while tapping the body or play the troublesome note until the feedback begins. The fi lter will automatically identify and eliminate the feedback. The “1” in the display will now light solid.3. At this point, you may continue to turn up your Volume as in step 2 to identify up to two more problematic resonances. Each is indicated via a fl ashing “2” or “3” during the search, in turn lighting solid when the resonance has beenidentifi ed.4. You may press the Volume knob at any time to cancel the search. The circuit will hold the notched frequenc(ies) in memory until the process is repeated.9Edit Mode ControlsPress Edit knob to enter Edit mode; the tuner’s green in-tune LED will light solid. Once in Edit, press the Edit knob repeatedly to step through the parameters. Each parameter is displayed using a single letter to represent its function.Adjust its value by turning the Edit knob. A number is displayed and the tuner’s sharp/fl at LEDs indicate positive/negative values.Note: F1 Aura is programmable and automatically saves your settings.To exit Edit mode, wait 10 seconds for the display to go dark, or press + hold the Edit knob for 2 seconds. You may also immediately exit by simultaneously pressing both the Edit and Volume knobs.10Edit Mode Controls (cont.)Image SelectF1 Aura is factory loaded with Images created especially for this instrument. Each Image corresponds to a different microphone type and position. Contact the guitar’s manufacturer to identify the microphone associated with each Image. Pickup EQBass, Mid and Treble controls allow you to fi ne tune the pickup signal. The pickup tone controls are designated by the capital letters T, M, B, corresponding to Treble, Mid, and Bass respectively.BlendDescribed in detail on page 8, Blend is represented by the letter X.12CompressorThe Compressor (C) parameter adjusts several settings within a sophisticated automatic leveling circuit. As you increase the value, your overall playing dynamics become increasingly limited, making softer notes louder while controlling loud spikes in your playing. This can be helpful in performances where you desire a more even level to your playing. At its maximum setting, there may be some overall increase in the output level.Anti-FeedbackThis parameter, indicated with the letter A, allows you to temporarily disable the automatic Anti-Feedback fi lter if desired. O = Off, I = On. See page 8 for details on how to use the Anti-Feedback circuit.13Edit Mode Controls (cont.)Image EQYou can program unique EQ settings for each of the Images. Unlike other Edit mode parameters, unique tone settings are saved with each Image and recalled when an Image is selected. In order to prevent dramatic or unwanted changes, the Image EQ’s Treble, Mid, and Bass controls are separated from the Pickup EQ and located “under” the Volume knob in Edit mode. They are identifi ed by a lowercase t, m, b, corresponding to Treble, Mid, and Bass respectively.To EQ an Image:1. Adjust the Blend control so that you are hearing 100% Image (page 8)2. Press Edit knob to enter Edit mode and select an Image3. Press the Volume button to select the Image Treble EQ (t)4. Turn the Edit knob to boost or cut the Image Treble EQ5. Repeat steps 3 & 4 to adjust the Image Mid (m) and Image Bass (b)14PowerPlug in the guitar, and F1 Aura switches on. To conserve the battery, remove the instrument cable from the guitar when the system is not in use.The tuner display will fl ash at power-up to tell you the preamp is on.Low Battery IndicatorWhen the tuner fl ashes “L” once every three seconds, you have approximately 1.5 hours before the battery is exhausted. Change it at the next opportunity.16Restore Factory Defaults1. Press + hold the Edit knob while plugging in the guitar. Continue to hold the Edit knob down when the tuner displays an “R.”2. Continue to hold the Edit knob, then press the Volume knob. Release both knobs. Factory reset is complete when the “R” stops fl ashing and the unit returns to normal operation.Defaults:• EQ for all Images reverts to fl at• Pickup EQ reverts to fl at• Blend is set to 50/50• Compressor is set to minimum• Anti-Feedback frequency is reset to 100Hz• Image selector reverts to Image #11718Electrical Specifi cationsDigital Signal Path:A/D, D/A conversion: 24-bitSignal Processing: 32-bitTypical in-use current consumption @ 9VDC: 18mATypical 9V lithium battery life: 54 hours Typical 9V alkaline battery life: 27 hours Nominal output impedance: 1k Ohm Recommended load impedance: 10k Ohm and up Maximum output level (onset of clipping): +5dBVBaseline noise: -92dBV Dynamic Range: 97dBBass control: ±12dB @ 70Hz Midrange control: ±12dB @ 1kHz Treble control: ±***********All specifi cations subject to change without notice.19。

新英汉摄影技术词典

新英汉摄影技术词典

新英汉摄影技术词典前言随着摄影技术的不断发展,摄影领域的专业术语也日新月异。

本词典旨在帮助读者更好地了解摄影技术术语,并在实践中运用。

无论是新手摄影爱好者,还是专业摄影师,都可以在本词典中找到所需的信息和参考资料。

A1. Aperture (光圈)光圈是调节镜头光线的大小,通常用f值表示。

较小的f值意味着更大的光圈,更多的光线进入镜头;而较大的f值意味着较小的光圈,限制光线进入镜头。

2. Auto Focus (自动对焦)自动对焦是指相机能够自动识别并对焦于所选择的目标,而不需要手动调整镜头。

B1. Bokeh (虚化)虚化是指在摄影中通过控制焦距和光圈大小,使得背景模糊,从而突出主体目标。

2. Bracketing (曝光补偿)曝光补偿是指通过连续拍摄同一景物的多张照片,每张照片的曝光参数有所不同,以确保至少有一张符合要求的曝光。

C1. Depth of Field (景深)景深是指摄影作品中清晰焦点与模糊焦点之间的范围,主要受光圈大小、镜头焦距和拍摄距离等因素的影响。

2. Cropping (裁剪)裁剪是指在后期处理中对照片进行裁剪或修剪,从而改变画面的构图和内容。

D1. Dynamic Range (动态范围)动态范围是指相机传感器能够捕捉的亮度范围,通常以光线最暗部分和最亮部分之间的差异来衡量。

2. Double Exposure (双重曝光)双重曝光是指在同一张底片或传感器上连续曝光两次或以上,从而叠加多个场景,创造出独特的效果。

E1. Exposure (曝光)曝光是指相机传感器或胶卷对光线的暴露时间,通过控制快门速度、光圈和ISO等参数来调节曝光量。

2. Electronic Viewfinder (电子取景器)电子取景器是指使用数字信息显示相机镜头捕捉画面的一种取景方式,通常在无反光镜相机中使用。

F1. Fisheye Lens (鱼眼镜头)鱼眼镜头是一种超广角镜头,其视角非常广阔,画面呈现出凸起状,造成鱼眼效果。

Acoustic Camera手册说明书

Acoustic Camera手册说明书

CHALLENGEResidents living in Welzow near the mining bridge were complaining about environmental noise. In order to reduce noise pollution, the acoustic sources of the bucket-wheel excavator needed to be localized. SOLUTIONThe Acoustic Camera as a mobile system allows high flexibility for outdoor measurements. Fast measurement and quick analysis enables users to easily identify sources as well as additional measurement positions of interest. With complete measurement set-up and postprocessing of the data, it takes only a few hours to completely identify the loudest sources on large machinery. The sound sources on the excavator can be precisely identified without stopping its operation. Once these sound sources are localized, they can be eliminated to reduce the effects of noise pollution.gfai tech GmbH Volmerstraße 3 12489 Berlin | Germany Ph.: +49 30 814563-750Fax: +49 30 814563-755E-Mail:****************www.gfaitech.de2D Outdoor MeasurementTHE ACOUSTIC CAMERA FOR ENVIRONMENTAL NOISE ANALYSIS AND HEAVY EQUIPMENTBENEFITS• Fast and easy set-up• No expensive machine downtime• Long distance measurement capability• Mobile system for flexible measurement locations• Advanced algorithms for precise localizationMEASUREMENT Measurement Object Microphone Array SoftwareData Acquisition Strip mining excavator Star48 AC Pro NoiseImage 4 Acoustic Photo 2D Recorder Spectral Photo Advanced Algorithms Data Recorder mcdRecThe Acoustic Camera was set up as a mobile system for full flexibility using a Star Array with 48 microphone channels. After an overview measurement from 150 meters away, the acoustic hot spots were localized. After the quick overview analysis, additional measurements were conducted closer to the sound sources for a more detailed analysis.www.gfaitech.de Page 2/2RESULTFor a reliable sound analysis, background noise was minimized by applying A-weighting to all data. In the first measurement from 150 meters distance, the main sources could be easily identified. In Fig. 1, the powerhouse and the derrick jib clearly stand out as the loudest areas.The second measurement was conducted closer to the derrick jib. As result of this measurement, the redirection point of the jib was identified as one of the loudest sources. A third measurement was made from a position on the bank with direct view to the jib. Several additional measurements were conducted from that position. A detailed analysis in the frequency domain was conducted using advanced algorithms in NoiseImage.The analysis using the HDR algorithm (High Dynamic Range) shows different sources and reflections on the ground within a dynamic of 35 dB(A). In the spectrogram (Fig. 3), the “banging” sounds are clearly visible. In order to find their location, these areas are simply marked as basis for the calculation region of the acoustic photo. The sound sources can be precisely identifiedas showninFig. 4. The broadband rattle sounds were produced by theshovel guidesimpacting the guide blockjust below the point of redirection.Fig. 1, left: powerhouse and derrick jib,right: derrick jibFig. 2direct view of the derrick jibFig. 3 Acoustic photo and spectrogram with a marked frequency range of 700 Hz - 1,1kHzFig. 4Acoustic photo and spectrogram withE-Mail:****************。

纹理物体缺陷的视觉检测算法研究--优秀毕业论文

纹理物体缺陷的视觉检测算法研究--优秀毕业论文

摘 要
在竞争激烈的工业自动化生产过程中,机器视觉对产品质量的把关起着举足 轻重的作用,机器视觉在缺陷检测技术方面的应用也逐渐普遍起来。与常规的检 测技术相比,自动化的视觉检测系统更加经济、快捷、高效与 安全。纹理物体在 工业生产中广泛存在,像用于半导体装配和封装底板和发光二极管,现代 化电子 系统中的印制电路板,以及纺织行业中的布匹和织物等都可认为是含有纹理特征 的物体。本论文主要致力于纹理物体的缺陷检测技术研究,为纹理物体的自动化 检测提供高效而可靠的检测算法。 纹理是描述图像内容的重要特征,纹理分析也已经被成功的应用与纹理分割 和纹理分类当中。本研究提出了一种基于纹理分析技术和参考比较方式的缺陷检 测算法。这种算法能容忍物体变形引起的图像配准误差,对纹理的影响也具有鲁 棒性。本算法旨在为检测出的缺陷区域提供丰富而重要的物理意义,如缺陷区域 的大小、形状、亮度对比度及空间分布等。同时,在参考图像可行的情况下,本 算法可用于同质纹理物体和非同质纹理物体的检测,对非纹理物体 的检测也可取 得不错的效果。 在整个检测过程中,我们采用了可调控金字塔的纹理分析和重构技术。与传 统的小波纹理分析技术不同,我们在小波域中加入处理物体变形和纹理影响的容 忍度控制算法,来实现容忍物体变形和对纹理影响鲁棒的目的。最后可调控金字 塔的重构保证了缺陷区域物理意义恢复的准确性。实验阶段,我们检测了一系列 具有实际应用价值的图像。实验结果表明 本文提出的纹理物体缺陷检测算法具有 高效性和易于实现性。 关键字: 缺陷检测;纹理;物体变形;可调控金字塔;重构
Keywords: defect detection, texture, object distortion, steerable pyramid, reconstruction
II

039_paper_shcherback_pixel_scan_crosstalk

039_paper_shcherback_pixel_scan_crosstalk

CMOS IMAGE SENSOR SENSITIVITY IMPROVEMENT VIA CUMULATIVE CROSSTALK REDUCTION Igor Shcherback1, Elad Gan1, Lior Blockstein2, Orly Yadid-Pecht1,21. Pixel-Scan, Ltd,114 Midreshet Sde-Boker, 84990, Israelshcigor@eladgan@Phone: 972-8-6255888/1412. The VLSI Systems Center, Ben-Gurion University,P.O.B 653, Beer-Sheva 84105, Israelblockstein@oyp@ee.bgu.ac.ilPhone: 972-8-6477257Fax : 972-8-6477620Digest:The global distortion that occurs in CMOS Image Sensors (CIS) is manifested in information leakage from each and every physical pixel to its surroundings or crosstalk (CTK) that might be photonic (e.g., optical, chromatic), electronic or both. This leakage degrades the spatial frequency, reduces overall sensitivity, color separation, contrast, and leads to additional noise after the color correction procedure, which results in failed products. The problem becomes worse as the pixel pitch shrinks.We have developed and present a generic technique for CIS cumulative CTK reduction which can be applied to any photosensors array situated on the common substrate.In the proposed approach the captured signal is reoriented internally such that the photocarriers captured by the "wrong" pixels are "restored" to the pixel they initially originated from, and the “improved” image is re/constructed without additional signal loss. The developed generic technique suits both color and monochrome CISs inherently considering optical as well as electronic CTK in the array.The developed methodology considers the CTK shares existing within the investigated pixel array as “known” (e.g., acquired by direct measurements by the spot light stimulation in three different wavelengths (e.g., ged, green, and blue), covering the visible spectrum), and enables to improve the overall sensor performance, resolution and sensitivity, i.e., improve its color sensitivity (separation/tint), increase quantum efficiency, eliminate “green disparity” (the mutual disparity of the signals obtained from the “green” pixels situated adjacent to the “blue” and “red” pixels correspondingly), increase effective sensor resolution (enlarge effective pixel count, increase sharpness, contrast, reduce blooming, smear, etc). Moreover, the presented method does not amplify and/or add signal noise, while inherently reducing the imager system noise, it improves overall sensor SNR.The proposed methodology was verified using a commercial advanced 2.2um pitch CIS camera, and is presented here as an example of the described technique applicability. Fig. 1 emphasizes the image quality and color rendering improvement achievable by the proposed method, as well as sensor SNR improvement. . Fig. 2 and Fig. 3 show the significant effective resolution improvement, such that details (spatial frequencies) that are almost undistinguishable in the original pictures can be easily resolved after the commulative CTK reduction. Fig. 2 also shows the improvement of the overall sensor sensitivity and contrast, blur reduction, etc. Fig.3 contains the evidence of the achievable color sensitivity improvement.Note that current work is still in progress, and we intend to introduce additional results confirming the presented technology applicability.Fig. 1: Example of the proposed technology applicability; Color/Tint rendering Improvement: Left – image obtained from the sensor; Right - same image after the proposed crosstalk reduction. The same demosaicing method was used for both images. Note the reduction of the spatial noise viewable at the left (and caused mainly by the “green disparity”, which is eliminated). The relative ∆E CIELab improvement (representing the improved color representation abilities of the sensor after CTK reduction) comes up to ~50% for certain colors.204060801001201401601802040608010012014016018020050100150200250204060801001201401601802040608010012014016018020050100150200250204060801001201401601802040608010012014016018020050100150200250Corrected imageFig. 2: Example of the proposed technology applicability; Left upper image - presents the original image captured by the sensor: Right upper image – after CTK reduction; bottom image shows the difference between the two. Picture sharpness is improved; such that the details that are almost undistinguishable in the original picture can be easily resolved after the CTK reduction. Note that the main difference concentrates "around the text", i.e., in the region where the camera resolution is damaged by CTK for the relatively high spatial frequencies range.Picture sharpness and contrast are alsoimproved.Fig. 3: Example of the proposed technology applicability: Left – circular diffractive-like image used for the sensor resolution checkup (aliasing is caused by the image shrink); Right - the normalized cross-sections comparison, i.e., from the captured image, and after CTK reduction. The intensity attenuation caused by the resolution fall with the spatial frequency (i.e., MTF) improvement after crosstalk reduction is evident especially for the higher frequencies range. Without loss of generality, the current example introduces the sensitivity and resolution improvement only for the RED color plane (i.e., after the same demosaic process was applied for both original and crosstalk-reduced images), such that higher signal obtained after CTK reduction correlates to the improved RED response. Similar results were obtained also for the Green and Blue sensor responses.。

Amersham Imager 600

Amersham Imager 600

GE HealthcareLife SciencesData file 29-0981-07 AA Imaging systems, software, and accessories Amersham™ Imager 600Amersham Imager 600 series is a new range of sensitiveand robust imagers for the capture and analysis of highresolution digital images of protein and DNA samples in gelsand membranes. These multipurpose imagers bring highperformance imaging to chemiluminescence, fluorescence,and colorimetric applications. The design of AmershamImager 600 combines our Western Blotting applicationexpertise with optimized CCD technology and exceptionaloptics from Fujifilm™. The system has an integrated analysissoftware and intuitive workflow, which you can operatefrom an iPad™ or alternative touch screen device, togenerate and analyze data quickly and easily.Amersham Imager 600 delivers:• Intuitive operation: You can operate the instrument froma tablet computer with an intuitive design and easy-to-useimage analysis software. You do not need prior imagerexperience or training to obtain high-quality results.Use the automatic capture mode for convenient exposure• Excellent performance:The system uses a super-honeycomb CCD and a large aperture f/0.85 FUJINON™ lens, which consistently delivers high-resolution images, high sensitivity, broad dynamic range (DR), and minimal cross-talk• Robustness:Combining minimal maintenance with our proven expertise in Western blotting and electrophoresis makes the imager well suited for multiuser laboratories. Amersham Imager 600 is an upgradable series of imagers that can grow with your imaging needs DescriptionAmersham Imager 600 series is equipped with a dark sample cabinet, a camera system, filter wheel, light sources, anda built-in computer with control and analysis software. Network connection and USB ports are standard (Fig 2). Fig 1. Amersham Imager 600 series is a range of robust and easy-to-use systems for chemiluminescent, colorimetric, and fluorescent image capture. Settings such as focus, filter, illuminator, and exposure type are automatically controlled by the integrated software. You would obtain high resolution images and precise quantitation of low signals with the multipurpose 16-bit3.2 megapixel camera fitted with a large aperture lens. The detector is cooled to reduce noise levels for high sensitivity and wide dynamic range. Rapid cooling leads to a short startup time, which makes the instrument ready to use in less than 5 min.You can place the sample tray at one of two different heights in the sample compartment to produce image-acquisitionareas of 220 × 160 mm and 110 × 80 mm, respectively.2 29-0981-07 AAThe system can be used for a wide range of applications and it is fully upgradable between four different configurations (Table 1). Each configuration can be used for chemiluminescent detection and nonquantitative gel documentation. The different configurations are equipped with light sources and filters for UV and white light trans-illumination, and red, green and blueTable 1. Amersham Imager 600 series comprises four different configurations. Amersham Imager 600 QC is designed for QC applicationsAmersham Imager 600Amersham Imager 600UVAmersham Imager 600RGB Amersham Imager 600QCWhite light (epi)××××Chemiluminescence ××××UV Fluorescence o ×××RGB Fluorescence o o ×o White light (trans) calibrated OD measurements oo××× StandardO OptionalFig 2. Amersham Imager 600 is ready to capture images within 5 min after startup. The imager is equipped with USB ports and a network connection.Network connectionUSB portsUSB ports TouchscreenPower switch Sample tray Cabinet doorepi-illumination for multiple fluorescence detection. Optical density (OD) measurements are calibrated for quantitation of colorimetric staining applications.You can operate the system via an integrated computer, controlled from a wireless iPad, a USB-connected touch screen, or a traditional monitor with a mouse and keyboard.Imaging performanceA bright, wide aperture FUJINON f/0.85 lens developed for chemiluminescent imaging projects sharp images onto a specially patterned CCD (Fig 3).Fig 3. The special octagonal interwoven pixel layout offers a dense matrix fora more efficient capture of light compared to a standard, square-pixel layout.Intuitive operation and analysisAmersham Imager 600 can be controlled from either an iPad or an alternative touch screen device. The user interface is intuitive and the workflow is easy to follow. The system is fully automated, which means that after startup, you do not need to perform focusing, insertion of light sources, changing of filters, calibrations, or other adjustments.When the system is in automatic image capture mode, it performs a short pre-exposure of the whole sample to determine the optimum exposure time for the strongest signal without saturating the image so that an accurate quantitation of the sample can be attained. In semi-automatic image capture mode, an automated exposureis made based on an area of interest defined by you. Exposure times are also easy to set manually.After image acquisition, the seamless workflow allows youto detect and quantitate bands, determine molecular weight, and perform normalization. The results are presented in both tabular and graphical formats so that you can easily and quickly analyze your data. For additional flexibility during data analysis, we offer ImageQuant™ TL software.You can use the system to obtain images of colorimetric markers and stains, such as Coomassie™ Blue or silver. Moreover, white light imaging can be combined with chemiluminescence and fluorescence imaging to generate overlay images of marker and sample. This feature allows quick molecular weight estimation and simplified documentation.The images can be stored in the system, on a USB memory stick or external hard drive, or in a network folder. Examples of imaging applicationsThe following examples of applications illustrate the performance and flexibility of Amersham Imager 600. Chemiluminescent Western blotting detection Quantitative Western blotting requires a signal response that is proportional to the amount of protein. A broad dynamic range with linear response allows you to simultaneously quantitate both high and low levels of proteins. The combination of Amersham Imager 600 with either Amersham ECL™ Prime or Amersham ECL Select™ resultsin a limit of detection in the picogram range and a dynamic range covering three orders of magnitude.Amersham Imager 600 has high sensitivity, which allowsyou to detect very weak signals in chemiluminescence andfluorescence applications for both protein and nucleic acids.Moreover, the wide dynamic range of the imagers—over fourorders of magnitude—allows weak and strong signals tobe quantitated accurately at the same time. The camera iscooled to -25°C to reduce dark noise giving less backgroundnoise during longer exposure times, which is especiallyimportant for the precise quantitation of very weak signalsin chemiluminescent Western blotting. The images areautomatically corrected for both geometric and intensitydistortion (radial, dark frame, and flat frame) in each imagingmode. This provides images that need minimal post-processingfor publication.RobustnessAmersham Imager 600 is a highly robust series of instruments,making it suitable for multi-user environments. The imagersdo not require calibration. Short exposure times and a fastanalysis workflow means that several researchers can use thesystem in the course of a day. The camera system is designedfor simple operation.29-0981-07 AA 34 29-0981-07 AAFig 6. The chemiluminescence mode allows simultaneous imaging of chemiluminescent samples and colored molecular weight markers. This image was taken from experiments for optimizing the expression of the protein DHFR in E. coli grown under different conditions.Fig 7. Evaluation of linearity, dynamic range, and limit of detection forfluorescence detection with Amersham Imager 600. A two-fold dilution series of phosphorylase b prelabeled with Cy5 shows a dynamic range of 3.3 orders of magnitude.Fluorescent imagingAmersham Imager 600 combined with Amersham ECL Plex™ provides high-quality data in applications that demand high sensitivity over a wide dynamic range. Furthermore, theminimal crosstalk of Amersham Imager 600, and the spectrally resolved dyes Cy™2, Cy3, and Cy5, makes it a suitable system for a wide range of multiplexing applications, such as the detection of several proteins at the same time or different proteins of similar size.Sample:E. coli lysateMembrane: Amersham Hybond ECL Blocking: 3% BSA in PBS-TMarker:Full range ECL Plex Fluorescent Rainbow Marker Primary antibody: Rabbit anti DHFR C-terminal 1:1000Secondary antibody: ECL Anti-rabbit IgG horseradish peroxidase 1:100 000Detection: Amersham ECL Select Imaging:Amersham Imager 600Imaging method: Chemiluminescence with colorimetric markerSample: Two-fold dilution series of LMW marker with Phophorylase bstarting at 200 ng Prelabeling: Cy5Imaging:Amersham Imager 600Imaging method: Fluorescence Cy5Limit of detection: 98 pg phosphorylase b Dynamic range:3.3 orders of magnitude123456789DHFR8.07.06.05.04.03.0Log protein amount (pg)L o g i n t e g r a t e d i n t e n s i t yPhophorylase b200 ng98 pgFig 5. Evaluation of limit of detection with Amersham Imager 600 forchemiluminescence using a two-fold dilution series of transferrin from 625 pg.8.07.06.05.04.03.0Log protein amount (pg)L o g i nt e g r a t e d in t en s i t yTransferrin625 pg2.5 pgSample : Two-fold dilution series of transferrin from 625 pg to 2.5 pg Membrane : Amersham Hybond P Blocking :3% BSA in PBS-TPrimary antibody : Rabbit anti-transferrin 1:1000Secondary antibody : ECL Anti-rabbit IgG horseradish peroxidase 1:75 000 Detection : Amersham ECL Select Imaging :Amersham Imager 600Imaging method :Chemiluminescence Limit of detection (LOD): 2.5 pg transferrinFig 4. A two-fold dilution series of NIH/3T3 cell lysate starting at 5 µg total protein was subjected to chemiluminescent Western blotting and ERK was detected with Amersham ECL Select. Dynamic range and linearity weredetermined. ERK could be detected in a cell lysate with 9.8 ng of total protein. Amersham Imager 600 showed a linear response for chemiluminescent detection with low noise, high sensitivity, and a wide dynamic range.8.07.57.06.56.05.55.04.54.0Log cell lysate amount (ng)L o g i n te g r a t e d i n t en si t yNIH/3T3 cell lysate5 µg9.8 ngSample: NIH/3T3 cell lysate two-fold dilution series starting at 5 µg Membrane : Amersham Hybond™ P Blocking : Amersham ECL Prime blocking agent 2% in PBS-T Primary antibody : Rabbit anti-ERK1/2 1:10 000Secondary antibody : ECL Anti-rabbit IgG horseradish peroxidase 1:100 000Detection : Amersham ECL Select Imaging : Amersham Imager 600Imaging method : Chemiluminescence Dynamic range: 2.7 orders of magnitude Amersham Imager 600 offers chemiluminescence imaging with an automatic overlay function. This allows simultaneous imaging of a chemiluminescent sample and a colored molecular weight marker. The overlay image retains the marker color.29-0981-07 AA 5Fig 8. Multiplex detection of total protein and target protein with Amersham ECL Plex and Amersham Imager 600. Detection of DHFR (Cy3 green) in nine different samples from a growth optimization of E. coli. Total protein in the samples was prelabeled with Cy5 (red). The overlay image shows the DHFR band in yellow. Crosstalk between Cy5 and Cy3 was minimal for Amersham Imager 600, which makes it suitable for multiplex applications.Fig 9. (A) Proteins stained with Coomassie Brilliant Blue and detected with Amersham Imager 600. The illustration shows nine different samples of E. coli lysates, from a growth optimization experiment for the expression of DHFR. Purified DHFR was used as a reference (sample 10). (B) Two-fold dilution series of the LMW-SDS Marker stained with SYPRO Ruby and detected with Amersham Imager 600.Sensitive imaging of total protein stainsProteins may be visualized by treating a gel with a total protein stain after performing 1D or 2D electrophoresis. The most commonly used stains are Coomassie Blue or silver staining. Fluorescent staining methods such as SYPRO™ Ruby protein gel stain have the advantage of being more sensitive.Sample: E. coli lysates Blocking:3% BSA in PBS-TPrimary antibody: Rabbit anti DHFR C-terminal 1:1000Secondary antibody: ECL Plex Goat anti rabbit-Cy3 IgG 1:2500Imaging:Amersham Imager 600Imaging method: Fluorescence Cy3, Cy5Sample: E. coli lysatesMarker:Full range ECL Plex Fluorescent Rainbow Marker Post staining: Coomassie Brilliant Blue Imaging:Amersham Imager 600Imaging method:Colorimetiric, white light epi-illuminationSample: Two fold dilution seires of LMW markerstarting at 1000 ng Post staining: Sypro RubyImaging:Amersham Imager 600Imaging method: Fluorescence Blue Epi excitation Limit of detection: 2 ng of carbonic anhydrase123456789Prelabeling Cy5Total proteinWB: DHFR Cy3Overlay12345678910Carbonic anhydraseLMW 1000 ng2 ng(A)(B)6 29-0981-07 AAFig 11. Image of a two-fold dilution series of LMW-SDS Marker in a gel stained with Coomassie Brilliant Blue. The image was recorded on Amersham Imager 600 in trans-illumination mode, which allows you to measure the optical density of protein bands without calibration.Sample:Two fold dilution series of LMW marker Post staining: Coomassie Brilliant Blue Imaging:Amersham Imager 600Imaging method: Colorimetric white transillumination Limit of detection: 16 ng of carbonic anhydrase Dynamic range:1.8 orders of magnitudeLog amount of carbonic anhydrase (ng)L o g i n t e g r a t e d i n t e n s i t yLMW markers1000 ngDNA imagingElectrophoretic separation of DNA is a common technique that is typically used for the analysis of vector cleavages, DNA purification, and verification of successful PCR. Traditionally, ethidium bromide (EtBr) has been used for visualizing DNA, but today there are many alternative DNA stains available, such as SYBR™ Green.Fig 10.Three-fold dilution series of KiloBase DNA Marker in agarose gel stained with SYBR Green and detected with Amersham Imager 600.Sample:Three-fold dilution series of KiloBase DNA Marker Post staining: Sybr Green I nucleic acid gel stain Imaging:Amersham Imager 600Imaging method: Fluorescence Cy2Limit of detection: 0.3 ng of total DNA Dynamic range:2.9 orders of magnitude8.07.57.06.56.05.55.04.54.0Log amount of total DNA (pg)L o g i n t e g r a t e d i n t e n s i t yKiloBase DNA Marker 250 ngQuantitative OD measurementAmersham Imager 600 QC is a dedicated configuration for densitometry applications in a QC environment. The system contributes to a reliable control of products because it is equipped with highly sensitive optics that can detect trace amounts of impurities accurately. Amersham Imager 600 QC is available with IQ/OQ and validation support.Amersham Imager 600 is autocalibrated for accurate and reliable measurements of optical density of proteins stained with colorimetric stains such as Coomassie or silver.Installation and Operational Qualification (IQ/OQ) validation servicesGE Healthcare offers validation services to support your equipment throughout its entire life cycle. Our validation tests and protocols are developed and approved byvalidation experts and performed by trained and certified service engineers. Our approach is in alignment with GAMP5, ICH Q8-10 and ASTM E2500, whereby validation activities and documentation focus on what is critical for end-product quality, and are scaled according to risk, complexity, and novelty. Our validation offering includes Installation and Operational Qualification (IQ/OQ), Requalification, and Change Control Protocols (CCP).29-0981-07 AA 7Ordering informationProductCode number Amersham Imager 60029-0834-61Amersham Imager 600UV 29-0834-63Amersham Imager 600QC 29-0834-64Amersham Imager 600RGB29-0834-67Accessories included (depending on configuration)Black tray AI60029-0834-17UV trans tray AI60029-0834-19White trans Tray AI60029-0834-18White Insert AI60029-0880-60Diffuser Board AI60029-0834-20Additional Accessories Gel sheets (for UV trans tray)29-0834-57Apple iPad 2 Wi-Fi –16GB –Black 29-0938-27Touch Screen Monitor with Stand 29-0939-66RangeBooster N USB Adapter 29-0928-76Additional SoftwareImageQuant TL 8.1, node locked license*29-0007-37ImageQuant TL 8.1, 5 x 1 node locked license*29-0008-10* External computer needed. Cannot be installed on Amersham Imager 600.Upgrade options Part/DescriptionRelevant for configurationCode number AI600 Upgrade 600 to 600 UV 60029-0834-22AI600 Upgrade 600 UV to 600 QC 600 UV 29-0834-24 AI600 Upgrade 600 QC to 600 RGB 600 QC 29-0834-25 AI600 Upgrade 600 UV to 600 RGB 600 UV29-0834-26IQ/OQ Validation service Amersham Imager 600 IQ/OQ29-0983-45Technical featuresTable 2. Amersham Imager 600 RGB specificationsCCD model:Peltier cooled Fujifilm Super CCD Pixel area 15.6 × 23.4 mmLens model:FUJINON Lens f/0.85 43 mm Cooling:Two-stage thermoelectric module with air circulation CCD Operating temperature -25°C Cooling down time:< 5 minDynamic range:16-bit, 4.8 orders of magnitude CCD resolution:2048 × 1472, 3.2 MpixelImage resolution:Maximum 2816 × 2048, 5.8 Mpixel Operation:Fully automated (auto exposure, no focus or other adjustment or calibration needed)Capture modes:Automatic, semi-automatic, manual (normal/incremental)Exposure time:1/10 s to 1 hourPixel correction:Dark frame correction, flatframe correction, and distortion correctionImage output:Gray scale 16 bit tif, Color image jpg, Gray scale jpg Sample size:160 × 220 mm Light sources:Blue Epi light: 460 nm Green Epi light: 520 nm Red Epi light: 630 nmUV transillumination light: 312 nm White light: 470 to 635 nmEmission filters:Cy2: 525BP20Cy3/EtBr: 605BP40Cy5: 705BP40Interface:USB 2.0 and Ethernet port Dimensions (W × H × D): 360 × 785 × 485 mmWeight: 43.6 kg (Amersham Imager 600 RGB)Input voltage:100 to 240 V Voltage variation:±10%Frequency:50/60 Hz Max power:250 W Operating temperature:18°C to 28°CHumidity:20% to 70% (no dew condensation)imagination at work GE, imagination at work, and GE monogram are trademarks of General Electric Company.Amersham, Cy, Hybond, ImageQuant, ECL, and ECL Select are trademarks of GE Healthcare companies.Coomassie is a trademark of Imperial Chemical Industries, Ltd. Fujifilm and Fujinon are trademarks of Fujifilm Corporation. iPad is a trademark of Apple Inc. SYBR is a trademark of Life Technologies Corporation. SYPRO isa trademark of Life Technologies Corp.CyDye: This product is manufactured under an exclusive license from Carnegie Mellon University and is covered by US patent numbers 5,569,587 and 5,627,027.The purchase of CyDye products includes a limited license to use the CyDye products for internal research and development but not for any commercial purposes. A license to use the CyDye products for commercial purposes is subject to a separate license agreement with GE Healthcare. Commercial use shall include:1. Sale, lease, license or other transfer of the material or any material derived or produced from it.2. Sale, lease, license or other grant of rights to use this material or any material derived or produced from it.3. Use of this material to perform services for a fee for third parties, including contract research and drug screening. If you require a commercial license to use this material and do not have one, return this material unopened toGE Healthcare Bio-Sciences AB, Bjorkgatan 30, SE-751 84 Uppsala, Sweden and any money paid for the material will be refunded.GE Healthcare UK LimitedAmersham PlaceLittle ChalfontBuckinghamshire, HP7 9NAUKGE Healthcare Europe, GmbHMunzinger Strasse 5D-79111 FreiburgGermanyGE Healthcare Bio-Sciences Corp.800 Centennial Avenue, P.O. Box 1327Piscataway, NJ 08855-1327USAGE Healthcare Japan CorporationSanken Bldg., 3-25-1, HyakuninchoShinjuku-ku, Tokyo 169-0073Japan29-0981-07 AA 02/2014For local office contact information, visit /contact /imagingGE Healthcare UK LimitedAmersham PlaceLittle ChalfontBuckinghamshire HP7 9NAUnited Kingdom。

图像拼接技术

图像拼接技术

王俊杰硕士研究生 研究方向:图像绘制 计算机视觉G 刘家茂硕士研究生 研究方向:图像处理G 胡运发博士生导师 研究领域:人工智能计算机视觉G 于玉教授G计算机科学2003Vol .30N.6图像拼接技术王俊杰刘家茂胡运发于玉(复旦大学计算机系上海200433)Research and Development of Image MosaicsWANG Jun -JieLIU Jia -Mao~U Yun -FaYU Yu(Department of omputer and Information Technology Fudan University Shanghai 200433)Abstract Image mosaics have been an active area of research in the fields of computer vision image processing andcomputer graphics in recent years .The automatic fast construction of unlimited field of view high -resolution image mosaics is a main research task of this area .According to the procedure of image mosaics the paper introduces and discusses image acguisition geometric corrections image register and image blending in detail .In the last part of the paper we make a discussion on some problems of research and point out the future research directiions .KeywordsImage mosaics Image acguisition Geometric corrections Image registration Panoramic image mosaicsImage blending1.引言图像拼接(Image Mosaics )是一个日益流行的研究领域 它已经成为照相绘图学~计算机视觉~图像处理和计算机图形学研究中的热点G 图像拼接解决的问题一般是 通过对齐一系列空间重叠的图像 构造一个无缝的~高清晰的图像 它具有比单个图像更高的分辨率和更大的视野G早期的图像拼接研究一直用于照相绘图学 主要是对大量航拍或卫星的图像的整合G 近年来随着图像拼接技术的研究和发展 它使基于图像的绘制(IBR )成为结合两个互补领域 计算机视觉和计算机图形学的研究焦点G 在计算机视觉领域中 图像拼接成为对可视化场景描述(Visual SceneRepresentations )的主要研究方法[1];在计算机图形学中 现实世界的图像过去一直用于环境贴图 即合成静态的背景和增加合成物体真实感的贴图G 图像拼接可以使IBR 从一系列真实图像中快速绘制具有真实感的新视图[2]G目前 一个特别流行图像拼接技术的应用是全景图的拼接 它是基于图像绘制虚拟现实场景创建和虚拟漫游的基础G 由于图像是独立拍摄的 在光滑表面上(如圆柱面[3*5]和球面[6*9])进行图像拼接对清晰度没有约束 并且可以避免不连续现象G 全景图提供一种在虚拟场景交互式浏览中良好的感觉 使用节点合成多个场景可以让用户在场景之间切换漫游[3 8] 利用计算机视觉的方法 能够从两个节点之间产生新的中间视点图像[4]G 与几何模型绘制真实场景相反 可以从这些节点重建场景的3D 几何模型[10]G图像拼接的其它一些应用有图像的稳定和变化检测[11] 图像分辨率增强[12] 视频处理[13] 包括视频压缩[14]和视频索引[15]G图像拼接过程一般分为四个步聚:(1)拍摄获取图像序列G 希望拼接的结果决定了图像获取方式G(2)镜头引起几何变形的校正G 利用图像序列数据或照相机模型进行校正G(3)对几何校正后的图像进行对齐G(4)图像的融合 消除对齐图像的不连续和缝隙G很多情况下 步聚(2)(3)是结合在一起进行的G 通常的图像拼接的技术也适用于全景图像拼接 因此本文将一同讨论全景图像拼接G本文首先介绍了图像获取(拍摄)的方法 接着讨论了图像几何变形的校正 然后对典型的图像对齐模型和对齐后的图像合成进行了详细的论述和比较分析G 最后指出了研究中存在的问题 以及今后的研究方向G2.图像获取(Image Ac !"isition )图像获取方法的不同导致取得输入图像的不同 最终拼接结果也不同G 图像获取由照相机拍摄时运动状态决定G 一般有三种情况[16]:(1)照相机固定在三角架上 旋转照相机拍摄G (2)照相机放置于一个滑辄上 平行移动照相机进行拍摄G (3)是一种普通的情况 人手持照相机 站在原地拍摄四周 或者沿着照相机的光轴垂直方向走动拍摄G2.1旋转照相机拍摄在这种情况下 放置照相机的三脚架在拍摄过程中一直在同一位置G 拍摄时 照相机绕垂直轴旋转 每旋转一定的角度 拍摄一张照片G 理想的情况下 照相机不绕照相机光轴旋转G拍摄得到一系列照片中相邻两张必须有部分重叠G 重叠区域大小是图像拼接最重要的影响因素G 文[3]建议相邻图像之间重叠比例达到50#G 重叠比例越大 拼接就越容易 但是需要的照片越多G旋转照相机拍摄由于照相机固定 不需要恢复过多参数 较容易实现G 但是 拍摄图像不在一个平面上 需要投影到同一个平面上 这将会导致图像质量下降G 一个解决方法是使用$141$短焦距9即广角镜头02.2平移照相机拍摄平移照相机指的是照相机在一个平行于成像平面的方向上平移0在固定焦距的情况下9照相机放置在一个滑轨上移动拍摄0物体和照相机的距离远近9或者拍摄物体的大小的变化9都会影响到最后的拼接结果0这种情况的缺点:拍摄的相片在一个平面上9全景图的三维感觉不如旋转拍摄的02.3手持照相机拍摄这种方法比较容易做到9手持照相机原地旋转拍摄9或按一定的路线平行于对象拍摄0但是9拼接手持照相机拍摄的照片是很困难的9因为在拍摄过程中9照相机的运动非常复杂0原地旋转拍摄类似于固定照相机旋转拍摄9但是角度控制9旋转控制都很差0沿一定路线移动时9类似于平移拍摄9控制距离和保持相同的成像平面很困难0为了减少这些影响9可以增加重叠比例9使照相机旋转角度~平移减小9因而减小相邻图像之间的不连续程度02.4存在的问题最常见的问题就是相邻图像之间光强的变化较大0理想情况下9相同的区域应该有相同的光强9但是因为光源变化或者照相机运动和光源平角的变化9导致光强的差异0另外一个和光条件相关的问题是反光区域9例如9镜子和闪亮金属0高亮光将会降低相应区域的对比度0场景中物体移动和拍摄时透镜引起的图像变形也将给对齐拼接带来困难0上述的方法一般限制了照相机的运动9但是实现中拍摄的图像存在小视差9不同比例的缩放和大角度旋转9这些都增加了对齐拼接的难度0因此很多文献[495917]要求照相机以最小运动视差旋转拍摄03几何校正(Geometric Corrections )几何变换通过重新映射9改变图像像素点之间空间关系0几何变换包括全局和局部的变换[18]0全局的变换通常有一个等式定义可以对整幅图像进行变换0局部变换只用于图像的部分9难以精确表达0几何变换通常用一些参数控制整个图像像素的运动9大量的像素信息对参数的估计很有效03.1照相机运动投影模型照相机运动投影模型包括2D 的仿射变换和透视变换0仿射变换可以表示为:f (I )=AI +b其中A 是变换矩阵9b 是平移矢量0矩阵A 控制仿射的缩放~旋转和剪切效果0透视变换是中心投影的射影变换9用非齐次射影坐标表达时9透视变换用8参数单应矩阵(homography )A 表示9改变矩阵A 进行透视校正9透视变换也表示为平面的分式线性变换:X/Y/T L 1W =m 11m 12m 13m 21m 22m 23m 31m 32T L 1 1I J T L 11I /=m 11I+m 12J+m 13m 31I+m 32J+1J /=m 21I+m 22J+m 23m 31I+m 32J+1其中W 为缩放参数0变换后的新图像坐标:I /=X //W 9J /=Y //W 0仅当图像之间没有运动视差时9相应图像的仿射变换才对所有像素有效0运动参数可以通过迭代求解[596912919]08参数单应矩阵能够精确描述照相机沿固定点旋转拍摄的不同视图间透视变换0具有代表性的透视变换有:投影变换9立体平面(Stereographic )变换和等距投影变换0这些变换有各自的特点:投影变换保持了线的形状9而立体平面变换保持了圆形状的特征0立体平面变换可以映射球面全部区域到一个投影平面上0等距投影变换可以看作球面的展平9它映射视图全部区域而不再是近似的情形03.2柱面~球面投影照相机绕固定点旋转拍摄获取图像序列的描述和合成一般用柱面和球面投影表示0在柱面~球面上合成的360度的图像称为全景图0全景函数[4](plenoptic function )描述了一个已知场景可能环境集合9全景图是从固定观察位置全景函数的一个采样0描述全景图通常采用柱面映射[3~5]9不考虑底面和顶面9在柱面坐标系统中用一致的采样合成全景图0球面映射[6~8]也作为描述全景图的环境0但球面全景图在极点处的异常使其构造显得复杂9极点附近的数值误差会导致自动拼接时评价误差不规则0在极点区域用鱼眼透镜获取图像和相对小尺寸图像会减轻上述异常的出现0在文[698]中9相邻图像间在映射到球面前的相对旋转运动可以避免图像对齐时的异常03.3照相机条扫描运动平面照相机在平移运动情况下9拍摄存在的视差会使几何校正变得复杂9一维照相机(如 pushbroom camera )扫描式的场景拍摄能够避免这种复杂的情况0这种方法能够用传统的照相机模拟9把两维图像序列中扫描条合并为一系列相邻区域0一维照相机可以旋转得到柱面映射图像或者平移得到正交映射图像[20]0文[21]中9两维图像的扫描条垂直于图像流0这些扫描条簇可以处理更多的运动9包括推进和光学缩放0文[22]提出了更复杂运动情况的描述0扫描条的图像获取记录了图像的范围和照相机的运动路径9这对场景的3D 完全重建显得非常有效[23]03.4任意几何校正的多项式变换多项式函数可以拟合图像和变换理想结果之间的对应关系9就像橡皮筋一样变换图像直到符合理想的结果0多项式变换通过一系列控制点的位移来定义图像空间变换9控制点将图像分成许多多边形区域9每个多边形区域使用双线性插值来填充非控制点04典型的图像对齐(Image Register )模型图像对齐是图像拼接过程的核心工作0其目标是找出对齐的两幅或多幅重叠图像之间的运动情况0早期9通过标定照相机或预先记录照相机的参数可以进行自动对齐9用户交互对图像对齐很有效0近年来9通过一些复杂的模型方法9实现了不需要用户交互的自动对齐算法0图像自动对齐方法一般分类如下:(1)在频域上操作的算法9例如基于FFT 的相位相关度法0(2)变换优化法9即迭代优化调整照相机参数9使评价误差最小0(3)基于图像特征的算法04.1相位相关度法(phase correlation )相位相关度法是一种简单的对齐方法9也是最有前途的自动对齐方法0它基于二维傅立叶变换的性质:空间域上的平241移等价于频域相位的平移,相位相关度法最早在1975年由Kuglin和~ineS提出,具有场景无关性,能够将纯粹二维平移的图像精确地对齐,De CaStro和Morandi发现用傅立叶变换确定旋转对齐,就象平移对齐一样,Reddy和Chatterji改进了De CaStro的算法,大大减少了需要转换的数量[24],两幅图像的平移矢量可以通过它们互功率谱(CroSS PoWer Spectrum)的相位直接计算出来,也就是说,相位相关是基于互功率谱的相位估计的,设两幅离散图像f1(:,y)和f2(:,y)在空间域简单的平移相关,f1(:,y)=f2(:-:0,y-y0)相应的傅立叶变换相关,F2(,7)=e-j(:0+7y0)F1(,7)采用两个图像规格化互功率谱时,相位相关度的公式如下,F1(,7)F e2(,7)F1(,7)F e2(,7)=e-j(:0+7y0)1FT8(:-:0,y-y0)规格化互功率谱的结果是简单复指数,即相位差,相位差的傅立叶反变换是在平移运动坐标上的脉冲,搜索最大值的位置就是两幅图像的对齐点,相位相关度法即使在有噪音情况下,如有重复的模式时,结果仍旧可靠和精确,为了加快计算速度,一般用快速傅立叶变换(FFT)计算,相位相关度法可以扩展到估计平面的旋转和缩放,利用傅里叶变换的旋转和缩放比例的性质,利用log polar坐标,将abs(F(,7))从直角坐标转换为极坐标(log(0),6),可以发现旋转和缩放在频域上相应的平移[24],文[25]用IDL (Interactive Data Language)实现了基于快速傅立叶变换的自动图像对齐算法,把直角坐标转换为log polar坐标,能够象平移一样描述旋转和缩放,从而确定旋转和缩放以对齐图像,文[26]中描述了一种基于Zernike momentS的快速有效的算法,这个算法允许更大程度地旋转~平移和缩放,相位相关度法一般需要比较大的重叠比例,通常要求对齐图像之间有50%重叠比例,文[27]中的实验在重叠比例为2.9%仍可以对齐时,讨论了重叠比例减少增加了位移脉冲函数的噪音,出现了许多对齐点相似于正确的对齐点,4.2变换优化法(Transf orm Optimisation)变换优化法是一种直观的图像拼接方法,这种方法潜在假设了两个空间图像通过一致性变换互相相关,通常情况下,变换矩阵的描述为3.1中的仿射变换或透视变换,变换优化是两个相关图像之间变换矩阵逐步求精的优化过程[5],可以运用Levenberg Marguardt最小化算法计算光强差平方和的最小值,设两幅图像I1,I2,变换优化矩阵M,变换等式为,X/= M X最小化目标评价函数如下,EV:,y(11(:,y)-12(:/,y/))2=E V:,y e2:,y目标评价函数不能区分两幅图小区域重叠,如果没有区域重叠,目标函数最小值可为0,为了纠正这个问题,文[27]重新定义了目标评价函数,EV:,y(11(:,y)-12(:/,y/))2N =EV:,ye2:,yN这个评价函数是基于重叠区域像素数量N的规格化,对于变换优化算法,很好的初始估计对于过程的收敛是非常重要的,通常有两个方法,相位相关度法(4.1)和高斯金字塔(分级匹配法),金字塔分级估计采用分级图像,分级求精的方案,先在低的(粗的)级别图像上计算确定粗略的结果,然后在高的级别图像计算求得精确的结果,分级估计处理包括,(1)金字塔构造G(2)运动估计G(3)图像变形G(4)逐步求精,但是实际情况中高斯金字塔初始估计可能不很理想[27],迭代速度太慢,甚至得不到对齐结果,而相位相关度法却是一个非常好的图像对齐初始化估计方法,4.B基于几何特征的图像对齐几何特征分为低级的特征,如边和角,和高级特征,如物体的识别和特征之间关系,基于低级特征对齐算法一般分为三步,首先过滤图像提取特征集,用这些特征集对两幅图像大概对齐,最后对转换迭代求精,文[2S]通过二维高斯模糊过滤可以得到一些低级特征模型,如边模型,角模型和顶点模型,因为角模型提供了比坐标点更多的信息,文[29]中利用几何角模型提出了图像对齐算法,文[30]中基于几何点特征优化匹配和文[31]中利用小波变换提取保留边(dege preServing)的视觉模型进行图像以齐,基于高级特征的图像对齐利用低级特征之间的关系或者通过识别出的物体实现对齐,文[32]利用特征图像关系图进行图像对齐,基于特征的图像对齐依赖特征的检测提取,如果特征不是很明显,可能会导致对齐失败,4.3.1角模型对齐(Corner Model)角模型用参数矢量(:0,y0,6,B,U,,)描述,其中,x0,y0,角的位置G G,角对称轴的方向G B,孔径角(aperture angle)G O,模糊程度参数G A,B,角内~角外灰度,角模型利用~arriS点和窗口尺寸初始化,得到在这个窗口内最适合这个模型的参数矢量和一种评测拟合质量方法,这种拟合方法是用图像灰度和模型内灰度平均最小平方差,取得平方差最小的值的角,这个模型拟合的角称为D.B.角,利用两个D.B.角中通过参数(:0,y0,6,B)定义的两条线计算变换优化的单应矩阵,这种方法比单纯的点特征计算单应矩阵减少了算法复杂性,在很少的D.B.角时,拟合角模型的步骤占了很大运算时间比例,为了提高算法的速度,使用了金字塔结构的方法由粗求精,在粗糙的等级上,相应的是用小的窗口进行拟合角,角模型对齐算法可以是对光轴的任意旋转和相当大的缩放,但是要求图像重叠的大约是50%,计算速度一般,4.3.2基于特征图像关系图的全局对齐由于对齐过程中会产生对齐误差累加,因此减小累加误差需要全局对齐,文[32]采用一种图描述图像序列时间空间相邻关系的2D拓扑结构,使用格点(grid point)而不是像素点组成拼接图像的镶嵌块,通过搜索图的最优路径进行全局对齐,图像关系图描述了图像序列之间的关系,节点表示全局对齐过程中的单个图像,如果两个节点对应的图像在空间或时间上是相邻的,那么这两个节点之间用一条边相连,空间相邻由两幅图像是否重叠决定,对于图像关系图中每一个节点,定义格点为特征点,它是图像对齐的根据,每个节点有一个格点链表,每个格点在其它节点链表中都有对应格点,图的拓扑描述可以搜索最优路径,这条路径把序列中每幅图像和参考图像连接起来,搜索是基于图中每条边的代价-341-函数0边的代价表示相邻两图像之间基于图像相关和格点变形的对齐程度0在最短路径上,累计的几何偏离和变形最小05图像融合(Image Blending)对齐的图像可能会存在强度或颜色的不连续和几何变形留下的缝隙0这可能是因为几何校正~动态的场景或光照条件的变化引起的,甚至是自动照相机,扫描设备的问题0这些问题可以通过图像融合来减小或消除0图像融合的方法很多,最简单的有光强平均和加权平均融合,复杂的有图像Vonoi权重法[33]和高斯样条(spline)插值法0其中文[5,34]的图像融合方法最为显著05.1采样权重函数文[5]提出了一种简单有效的方法,对合成结果的每个图像采样的权重分布这样定义,离图像中心越近的像素对最终合成结果贡献越大0即图像中心的权重大,边缘的小0权重分布函数表现为三角形,像一顶帽子,又称帽子函数(hatfunction)0利用帽子函数加权平均算法[35],C(I,y)=kz(I,y)1k(I,y)kz(I,y,k)z(I,y,k)=1-Iz c k-D12>1-yk -D12z(I,y,k)为帽子函数0z c 和是第k个图像的宽和高0这种方法可以完全消除边缘手工的痕迹,但是边缘低频的斑纹仍旧可见05.2基于欧氏距离的有效权重[34]图像的每个像素都分配权重,这个权重与到边缘(或最近不可见点)的距离成比例0在拼接时,主要目的是减少离边缘近的像素点的光强贡献0融合算法中计算距离映射d(X,y),利用块距离和欧氏距离,计算到最近的透明点(a=0)或边的距离0融合所有变形图像公式,C(I,y)=kz(c(I,y))1k(I,y)kz(c(I,y))z是单调函数,z(I)=I01k是第k幅变形图像的光强函数0c 的计算很简单,取离矩形四条边距离的最小值0文[34]算法边缘像素对拼接图像没有贡献,随着向图像中心的逐渐增加0文[5]算法中,边缘的点对拼接图像的贡献是50%0因此前者的算法减少了合成图像时光强颜色的不连续性0结论和未来的研究在本文中,我们论述和分析比较了图像拼接过中程每个步骤(图像获取~几何变形校正~图像对齐和图像融合)的主要方法模型0尽管这些方法模型已经比较完善,但是实际研究中还存在一些更复杂的问题,这也正是我们今后的研究方向0通常情况下,为了简化几何校正和对齐,图像获取时对照相机运动总是做了一定限制,但是这些限制简化不能描述实际中的照相机运动情况,如手持照相机的任意运动拍摄0不限制的照相机运动会增加几何校正和对齐的复杂性0在图像拼接时,可能会找不到一种合适的几何变形校正模型0一些不知道的图像之间的运动关系很难选择一个有效的校正模型,这会导致对齐结果中遗留不合适变形的痕迹,这个问题或许可以在融合中减弱或消除,但不是图像拼接的初衷0因此寻找一个合适的几何校正模型对对齐和融合是有利的0对于照相机任意的运动和对三位视差的检测和校正,需要建立一个通用(或近似通用)的照相机运动模型0在图像对齐中,希望能够找到一种更好的输入图像之间优化变换算法,不必要考虑不利的因素,如图像间光强跳变,不理想的照相机运动和缺少明显的特征0图像对齐算法的计算量一般都很大,在商业应用中可能是无法忍受的,因此拼接效果和计算量的折衷是商业应用必须要考虑的0既考虑了拼接结果的质量,又能兼顾到效率,这对算法的要求很高0通过本文的论述,我们希望能够提供给读者国际上图像拼接技术最新研究动态,并且对我国基于图像绘制技术的进一步研究有所启发和帮助0参考文献1Anandan P,et al.IEEE Workshop on Representations of Visual Scenes.Cambridge,Massachusetts,IEEE Computer Society Press,19952Kang S B.A survey of image-based rendering technigues, [Technical Report97/4].Digital Eguipment Corporation, Cambridge Research Lab,Aug.19973Chen S E.Guicktime VR-An image based approach to virtual environment navigation.In,Proc.of ACM SIGGRAP,19954McMillan L,Bishop G.Plenoptic modeling,An image-based rendering system.In,Proc.of ACM SIGGRAP,1995.39~465S Z eliski R.Video mosaics for virtual environments.IEEE Computer Graphics and Applications,March1996.22~306S Z eliski R,Shum Y.Creating full vie W panoramic image mosaics and environment maps.In,Proc.of ACM SIGGRAP, 19977X iong Y,Turko W ski K.Creating image-based VR using a self-calibrating fisheye lens.In,Proc.of IEEE Conf.on Computer Vision and Pattern Recognition,19978Coorg S,Master N,Teller S.Acguisition of a large pose-mosaic dataset.In,Proc.of IEEE Conf.on Computer Vision and Pattern Recognition,19989Coorg S,Teller S.Spherical Mosaics W ith Guaternions and Dense Correlation.I J CV,2000,37(3),259~27310Shum Y,an M,S Z eliski R.Interactive construction of3d models from panoramic mosaics.In,Proc.of IEEE Conf.on Computer Vision and Pattern Recognition,J une199811Morimoto C,Chellappa R.F ast3d stabili Z ation and mosaic construction.In,Proc.of IEEE Conf.on Computer Vision and Pattern Recognition,1997.660~66512Capel D,Z iserman A.Automated mosaicing W ith super-resolution Z oom.In,Proc.of IEEE Conf.on Computer Vision and Pattern Recognition,199813Irani M,Anandan P,Bergen J,Kumar R,su S.Mosaic representations of video seguences and their applications.Signal Processing,Image Communication,special issue on Image and Video Semantics,Processing,Analysis,and Application,1996,8(4)14Lee M-C,et al.A layered video ob ect coding system using sprite and affine motion model.IEEE Transactions on Circuits and Systems for Video Technology,1997,7(1),130~14515Irani M,Anandan P.Video indeXing based on mosaic representations.In,Proc.of the IEEE,1998,86(5),905~921(下转第150页)441(a D采集到的彩色图像(b D检测到的人脸区域图11部分试验结果用嘴唇这一特征进行特征匹配~找到最后的人脸区域O实验表明~此方法执行速度快~鲁棒性好~检测结果正确率高~且对人的头部运动适应性好O此方法适合于视频实时图像中的人脸检测与跟踪~该方法的进一步改进还可用于一般物体的实时检测与跟踪O参考文献1Yang G.~uman face detection in a complex background.Pattern Recognition~1994~27(1OD:53~632Brunelli R~Poggio T.Face recognition:feature versus template [J].IEEE Trans PAMI~1993~15(1OD:1O42~1O523Sung K~Poggio T.Example-based learning for view-based human face detecion.IEEE Trans Pattern Analysis and Machine Intelligence~1998~18(1D:39~514Rowley~A~Balnja S~Kanade T.Neural network-based face detection.IEEE Trans Pattern Analysis and Machine Intelligence~1998~2O(1D:23~285艾海舟~梁路宏~徐光佑~张钹.基于肤色和模板的人脸检测.软件学报~2OO1~12(12D:1784~179O6康学雷~邵凌~张立明.一种基于肤色和模板的人脸检测方法.红外与毫米波学报~2OOO~6:2O9~2147Chai D~Ngan K N.Face segmentation using skin-color in videophone application.IEEE Trans.CSVT~1999~9(4D:55O~566 8姚鸿勋~刘明宝~高文.基于彩色图像的色系坐标变换的面部定位与跟踪法.计算机学报~2OOO~2(2D:158~165(上接第144页D16Chen C-Y~Klette R.Image Stitching-Comparisons and New Technigues.CAIP~1999.615~62217Kang S B~Weiss R.Characterization of errors in compositing panoramic images.In:IEEE Computer Society Conf.on Computer Vision and Pattern Recognition(CVPR*97D~San Juan~Puerto Rico~1997.1O3~1O918Gumustekin S.A Introduction to Image Mosaicing.http: .tr/~July~199919Sawhney~S~Kumar R.True multi-image alignment and its application to mosaicing and lens distortion correction.In:Proc.of IEEE Conf.on Computer Vision and Pattern Recognition~ 1997.45O~4562O Peleg S~~erman J.Panoramic mosaics by manifold projection.In: Proc.of IEEE Conf.on Computer Vision and Pattern Recognition~199721Rousso B~Peleg S~Finci I.Mosaicing with generalized strips.In: DARPA Image Understanding Workshop~May199722Russo B~Peleg S~Finci I.Rav-Acha A.Universal mosaicing using pipe projection.In:Proc.of IEEE Intl.Conf.on Computer Vision~ 199823Rademacher P~Bishop G.Multiple-center-of-projection images.In:Proc.of ACM SIGGRAP~~199824Reddy B S~Chatterji B N.An FFT-based technigue for translation~rotation~and scale-invariant image registration.IEEE Trans.Image Processing~1996~5(8D:1266~127125Xie~ongjie~et al.Automatic image registration based on FFT algorithm and IDL/ENVI.In:Intl.Conf.on Remote Sensing and GIS/GPS-ICORG~2OO126Badra F~Oumsieh A~Dudek G.Rotation and Zooming in ImageMosaicing.In IEEE WACV~1998.5O~5527Richmond J.Image Mosaicing http:www.cs.ubc.ca/~ jlrichmo/~199928Blaszka T.Rachid Deriche Recovering and Characterizing Image Features Using An Efficient Model Based Approach:[Technical Report2422].INRIA~Nov.199429Zoghlami I~Faugeras O~Deriche ing Geometric Corners to Build a2D Mosaic from a Set of Images.In:Proc.of the Intl.Conf.on Computer Vision and Pattern Recognition~Puerto Rico~IEEE~19973O Nielsen F.Adaptive Randomized Algorithms for Mosaicing Systems.IEICE TRANS.INF.g SYST.~VOL.E83CD~2OOO (7D31Bao P~Xu plex Wavelet-based Image Mosaics using Edge-Preserving Visual Perception puters and Graphics~ 1999~23(3D:3O9~32132Kang E~Cohen I~Medioni G.A graph-based global registration for2D mosaics.In:Proc.of the15th Intl.Conf.on Pattern Recognition~Barcelona~Spain~2OOO33Sawhney~S~Kumar R G~Bergen J~Dixon D~Paragano V.Videobrush TM:Experiences with consumer Video Mosaicing.In IEEE WACV~1998.56~6234Shum~-Y~Szeliski R.Panoramic Image Mosaics~Microsoft Research:[Technical Report MSR-TR-97-23]35WONG~Chun Kit Wilbur YANG~Ruiduo Z~ANG~Yan.Image Mosaics under Simple Planar Scenes and its Application~A Final Report Submitted to The Course COMP641F Introduction to Image Processing at The~ong Kong University of Science and Technology~2OO1O51。

声学照相机

声学照相机

声学照相机系统ACOUSTIC CAMERA(用眼睛“听”声音)◎使用眼睛,技术人员可以更快更灵活的收集到所需信息。

目前磁共振成像技术(MRI)及红外照相机就用于提供人眼看不到的信息。

但是当考虑噪声问题时,目前仍然是一些常用的手段。

如果声音也能用眼看到会有什么好处呢?我们周围的世界用声学照相机来观测会是什么样呢?◎声学照相机系统(Acoustic Camera )是德国GFAL公司研制的用眼睛来“听”声音的声学测试系统,它是一种轻型的模块化的便携设备,用于对声源进行定位和分析。

逼真、准确和快速的声学虚拟成像的技术能够精确定位声源和指出质量问题。

◎Acoustic Camera应用的范围很大,有噪声的地方都可以应用,包括声学实验室,开放场合,恶劣的工厂噪声环境等。

基础的配置包括一组传声器阵列,数据采集器,笔记本电脑和配套软件"NoiseImage"。

不同的应用场合配备不同的阵列方式,阵列方式可以互换并配备高性能的数字照相机(1280×960像素),传声器32-36个可选。

◎数据采集设备为acoustic camera专配,多通道,大频率范围,结构紧凑,质量轻,并可以同时记录其他参数的模拟或者数字信号。

环形阵列(32)适用于声学实验室∙测量距离 0.7...5 m∙USB 照相机 (1280 x 1024像素), 定焦镜头∙图形范围100 Hz...20 kHz (50 kHz)∙传声器动态范围35...130 dB, 40 Hz...20 kHz, -1 dB, 24...120 dB, 30 Hz...20 kHz, 3 dB (LN)∙选择性 0,5 dB , 1 kHz∙传声器微软总线连接∙直径 75 cm, 重 5 kg星形阵列(36)适用于长距离车测试∙USB 照相机 (1280 x 1024像素), 定焦镜头∙图形范围100 Hz - 7 kHz (> 6dB)∙典型测量距离 3-300 m∙长 2 m (折叠后)∙重10 kg∙反向衰减-21 dB < 传声器>∙动态测量范围 35 - 130 dB, 30 Hz - 20 kHz (50 kHz)立方阵列(32)适用于全方位测量,一般用于小封闭空间高频测量∙内部空间适用∙分析破裂、滴答和其他噪声,风噪声∙立方体直径约35 cm, 重1kg∙测量距离 0.3...1.5 m∙声传播透明∙USB 照相机 (1280 x 1024像素), 定焦镜头∙图形范围 1 kHz...10 kHz∙阵列反向衰减 -20 dB∙传声器微软总线连接∙传声器动态范围 35...130 dB, 30 Hz...20 kHz (50 kHz) ∙选择性 0,5 dB, 1 kHz。

Acoustic camera

Acoustic camera

专利名称:Acoustic camera发明人:Dipen N. Sinha,John F. Brady申请号:US14398367申请日:20130313公开号:US10054676B2公开日:20180821专利内容由知识产权出版社提供专利附图:摘要:Apparatus for generating accurate 3-dimensional images of objects immersed in liquids including optically opaque liquids which may also have significant soundattenuation, is described. Sound pulses are caused to impinge on the object, and the time-of-flight of the reflected sound is used to create a 3-dimensional image of theobject in almost real-time. The apparatus is capable of creating images of objects immersed in fluids that are optically opaque and have high sound attenuation at resolutions less than about 1 mm. The apparatus may include a piezoelectric transducer for generating the acoustic pulses; a high-density polyethylene compound acoustic lens, a 2-dimensional segmented piezoelectric detecting array positioned behind the lens for receiving acoustic pulses reflected by the object, the electric output of which is directed to digital signal processing electronics for generating the image.申请人:LOS ALAMOS NATIONAL SECURITY, LLC地址:Los Alamos NM US国籍:US代理机构:Cochran Freund & Young LLC代理人:Samuel M. Freund更多信息请下载全文后查看。

利用场景光照识别优化的双目活体检测方法

利用场景光照识别优化的双目活体检测方法

DOI : 10.11992/tis.201912026利用场景光照识别优化的双目活体检测方法林峰1,杨忠程2,冯英2,颜水成2,魏子昆2(1. 贵阳市信捷科技有限公司,贵州 贵阳 550081; 2. 上海依图网络科技有限公司,上海 200051)摘 要:人脸识别是生物特征识别技术中应用最广的技术之一。

其中,能判断人脸图像是否是真实人脸的活体检测模块,是系统安全运行的重要保障。

目前从安全度和经济性两方面综合考虑,最常用的活体检测方法是双目活体检测。

但由于不同场景下光线亮度和角度变化很大,拍摄的人脸图片质量参差不齐,严重影响了活体检测的质量。

针对这一问题,提出了通过对场景光照识别进行优化从而提升检测准确度的双目活体识别算法。

算法通过串级PID 算法对摄像头的感光度和补光灯进行控制,并利用人脸识别算法定位优化测光区域,从而对不同的光线强度和角度采取不同的策略。

经过实验验证:本方法将活体检测在复杂场景下的准确率提升约30%,保证了算法在室内外不同光照场景下的有效性。

关键词:人脸活体检测;人脸防伪;展示攻击检测;身份识别;生物识别安全;深度学习;卷积神经网络;PID 控制中图分类号:TP391 文献标志码:A 文章编号:1673−4785(2020)01−0160−06中文引用格式:林峰, 杨忠程, 冯英, 等. 利用场景光照识别优化的双目活体检测方法[J]. 智能系统学报, 2020, 15(1):160–165.英文引用格式:LIN Feng, YANG Zhongcheng, FENG Ying, et al. Binocular camera based face liveness detection with optimized scene illumination recognition[J]. CAAI transactions on intelligent systems, 2020, 15(1): 160–165.Binocular camera based face liveness detection with optimizedscene illumination recognitionLIN Feng 1,YANG Zhongcheng 2,FENG Ying 2,YAN Shuicheng 2,WEI Zikun 2(1. Guiyang Xinjie Technology Co., Ltd., Guiyang 550081, China; 2. YITU Tech, Shanghai 200051, China)Abstract : Face recognition is one of the most widely applied biometric identification technologies, in which face live-ness detection aiming to determine whether a face is genuine or fake, is used to help face recognition systems defend against replay and print attacks, and thus ensure system security. Considering safety and economy, binocular camera based face liveness detection is most commonly adopted at present. However, due to significant variations in lighting conditions of different scenes as well as face poses, the captured face images are often of low quality, which greatly harms the performance of face liveness detection. In this paper, we propose a binocular camera based face liveness de-tection algorithm, which improves detection performance through optimizing scene illumination recognition. In particu-lar, the proposed algorithm uses the cascaded PID algorithm to adjust the light sensitivity and light supplement of the camera subject to specific lighting and pose angles. It also modifies the photometric range to be within the face area in the case of backlight to ensure effectiveness of the light exposure and supplement control strategy. Extensive experi-ments have been conducted and the results show that the proposed model outperforms other methods by around 30% in accuracy in complex scenes, with ensured generalizability to diverse application scenes.Keywords : face liveness detection; face anti-counterfeiting; display attack detection; identity recognition; biometric se-curity; deep learning; convolutional neural network; PID control随着科技的进步和社会的发展,生物识别技术被逐渐地应用到我们的日常生活中,比如人脸收稿日期:2019−12−20.通信作者:杨忠程. E-mail :****************************.第 15 卷第 1 期智 能 系 统 学 报Vol.15 No.12020 年 1 月CAAI Transactions on Intelligent SystemsJan. 2020识别、指纹识别、声音识别等。

imatest-Noise_in_photographic_images

imatest-Noise_in_photographic_images

imatest-Noise_in_photographic_images/doc/948e9573e518964bcf847cae.html /doc/948e9573e518964bcf847cae.html /docs/noise/ Noise in photographic imagesNoise in photographic imagesIntroductionNoise is a random variation of image density, visible as grain in film and pixel level variations in digital images. It is a key image quality factor; nearly as important as sharpness. Since it arises from basic physics— the photon nature of light and the thermal energy of heat— it will always be there. The good news is noise can be extremely low— often impreceptably low—in digital cameras, particularly DSLRs with large pixels (5 microns square or larger). But noise can get ugly in compact digital cameras with tiny pixels, especially at high ISO speeds.In most cases noise is perceived as a degradation in quality. But some Black & White photographers like its graphic effect: Many favor 35mm 35mm Tri-X film. The pointillist painters, most notably George Seurat, created “noise” (specks of color) by hand; a task that can be accomplished in seconds today with Photoshop plugins.But by and large, the majority of photographers, especially color and large-format photographers, dislike noise with good reason.Noise is measured by several Imatest modules. Stepchart produces the most detailed results, but noise is also measured in Colorcheck, SFR, and Light Falloff.Appearance(A) Const. sensor noise (B)Const.pixelnoise(C)Nonoise(D)CanonEOS-10DISO 1600The appearance of noise is illustrated in the stepchart images on the right. Noise is usually measured as an RMS (root mean square) voltage. The mathematics of noise is presented in a green box at the bottom of this page.The stepcharts in columns (A)-(C) are simulated. They are assumed to have a minimum density of 0.05 and density steps of0.1, identical to the Kodak Q-13 and Q-14. They have been encoded with gamma = 1/2.2 for optimum viewing at gamma = 2.2 (the Windows/Internet standard). Strong noise— more than you’d find in most digital cameras— has been added to columns (A) and (B). Column (C) is noiseless.The fourth column (D) contains an actual Q-13 stepchart image taken with the Canon EOS-10D at ISO 1600: a very ISO high speed. Noise is visible, but admirably low for such a high ISO speed (thanks, no doubt, to software noise reduction).The noise in (A) is constant inside the sensor, i.e., before gamma encoding. When it is encoded with gamma = 1/2.2, contrast, and hence noise, is boosted in dark areas and reduced in light areas. The Kodak publication, CCD Image Sensor Noise Sources, indicates that this is not a realistic case. Sensor noise tends to increase with brightness.The noise in (B) is uniform in the image file, i.e., its value measured in pixels is constant. This noise must therefore increase with brightness inside the sensor (prior to gamma encoding), and hence isf-stops (or EV or zones; a factor of two in exposure), i.e., referenced to the originalscene. Noise measured in f-stops corresponds closely to human vision. See f-stop noise,below.Noise as a simple (average) numberA good single number for describing noise is the average noise of the Y (Luminance)channel, measured in pixels (normalized scene density difference of 1.5), shown in thelower plot in the figures, below. The lightest and darkest zones, representing chart densitiesgreater than 1.5 or less than 0.1 are excluded from the average.S/N or SNR (dB) as a function of pixel level or exposure, where S is the pixel level of the individual test chart patch.S/N or SNR (dB) as a simple (average) number.a noise spectrum plot, described below.In addition, Light Falloff displays a spatial map of noise.Characteristic Stepchart results are shown below.Column (B): Uniform pixel noiseThis image has simulated uniform pixel noise(i.e., constant noise in the image file, measuredin pixels).The upper plot is the tonal response (orcharacteristic curve) of the camera. It shows theexpected ideal response for encoding withgamma = 1/2.2 = 0.4545: a straight line withslope = 0.4545 for the log-log plot.The middle plot shows noise measured in f-stops (or EV or zones). Noise increases moresteeply for dark regions (large negative values ofLog Exposure) in actual sensors.The lower plot shows noise measured in pixellevels, normalized to the difference in pixel levelsfor a density range of 1.5. It is relatively constant,showing only statistical variation, except for thebrightest level (on the right), where the noise isreduced because some samples are clipped atpixel level 255.The lower plot also contains the single numberused to characterize overall noise performance:the average Luminance channel noise (Y =5.59%, on the right near the top). This number isvery high; it corresponds to poor image quality.This is SNR in dB for simulated uniform pixelnoise. Because of the gamma = 2.2 encoding,SNR inproves by 6.02/2.2 = 2.74 dB for eachdoubling of exposure (0.301 density units);roughly 9.1 dB per decade (1 density unit).Column (D): Canon EOS-10D at ISO 1600 Unlike the above figure (for the image in column (B)), real data for the Canon EOS-10D at ISO1600 is used.The upper plot shows a slight tonal response “S”curve superimposed on the gamma curve (whichis a straight line in this log-log plot). Someconverter settings (such as “low contrast”) resultin a much more pronounced “S” curve.The middle plot shows f-stop noise, whichincreases dramatically in the dark regions.The lower plot shows the normalized pixel noise.It increases in the dark regions. This increase isdue in part to the high ISO speed. Digitalcameras achieve high ISO speed by amplifyingthe sensor output, which boosts noise,particularly in dark regions. This curve looksdifferent for the minimum ISO speed: noisevalues are much lower and subject to morestatistical variation.This is SNR in dB for the Canon EOS-10D atISO 1600. (This display option was introducedwith Imatest 2.3.16). SNR improves by about 6dB for each doubling of exposure (0.301 densityunits); roughly 20 dB per decade (1 density unit),which is what would be expected for constantsensor noise. (This curve would be dramaticallydifferent at lower ISO speeds.)Noise summaryThere are two basic types of noise.Temporal. Noise which varies randomly each time an image is captured. Measured byStepchart and Colorcheck using two identical input images.Spatial or fixed pattern. Noise cause by sensor nonuniformities. Sensor designers havemade heroic and largely successful efforts to minimize fixed pattern noise.Temporal noise can be reduced by signal averaging, which involves summing N images, then dividing by N. Picture Window Pro performs signal averaging using the Composite or Stack Images transformations, as described in Using Picture Window Pro in Astrophotography.When individual images are summed N times, the signal pixel level increases by N. But since temporal noise is uncorrelated, noise power (rather than voltage or pixel level) is summed. Since voltage is proportional to the square root of power, the noise pixel level (which is proportional to noise voltage) increases by sqrt(N). The signal-to-noise ratio (S/N or SNR) improves by N/sqrt(N) = sqrt(N). When four images are averaged, S/N is improved by a factor of 2.Several factors affect noise.Pixel size. Simply put, the larger the pixel, the more photons reach it, and hence the better the signal-to-noise ratio (SNR) for a given exposure. The number of electrons generated by thephotons is proportional to the sensor area (as well as the quantum efficiency). Noise power is also proportional to the sensor area, but noise voltage is proportional to the square root of power and hence area. If you double the linear dimensions of a pixel, you double the SNR.The electron capacity of a pixel is also proportional to its area. This directly affects dynamicrange.Sensor technology and manufacturing. The biggest technology issue is CMOS vs. CCD.We won’t discuss it in detail here. Until 2000 CMOS was regarded as having worse noise, but ithas improved to the point where the two technologies are comparable, differing only in detail.CMOS is less expensive because it is easy to add functionality to the sensor chip. Technology also involves other aspects of sensor design and manufacturing, all of which will improvegradually with time.ISO speed. Digital cameras control ISO speed by amplifying the signal (along with the noise) at the pixel. Hence, the higher the ISO speed the worse the noise. To fully characterize a sensor it must be tested at several ISO speeds, including the lowest and highest.Exposure time. Long exposures with dim light tend to be noisier than short exposures with bright light, i.e., reciprocity doesn’twork perfectly for noise. To fully characterize a sensor it should be tested at long exposure times (several seconds, at least). Digital processing. Sensors typically have 12-bit analog-to-digital (A-to-D) converters, sodigitization noise isn’t usually an issue at the sensor level. But when an image is converted to an 8-bit (24-bit color) JPEG, noise increases slightly. The noise increase can be worse (“banding”can appear) if extensive image manipulation (dodging and burning) is required. Hence it is often best to convert to 16-bit (48-bit color) files. But the output file bit depth makes little difference in the measured noise of (unmanipulated) files.Raw conversion. Raw converters often apply noise reduction (lowpass filtering) andsharpening (see Noise frequency spectrum, below), whether you want it or not; even if NR and sharpening are turned off. This makes it difficult to measure the sensor’s intrinsic properties.General commentsImatest subtracts gradual pixel level variations from the image before calculating noise (thestandard deviation of pixel levels in the region under test). This removes errors that could becaused by uneven lighting. Nevertheless, you should take care to illuminate the target as evenly as possible.The target used for noise measurements should be smooth and uniform– grain (in film targets) or surface roughness (in reflective targets) should not be mistaken for sensor noise. Appropriate lighting (using more than one lamp) can minimize the effects of surface roughness.Scene-referenced noiseIt is often valuable tomeasure noise relativetop the scene rather thanto the pixel levels. In doingso we need to take theresponse of the humaneye into account.The human eye respondsto relative luminancedifferences. That’s whywe think of exposure interms of zones, f-stops,or EV (exposure value),all of which correspond toa factor of two change inexposure.The eye’s relativesensitivity is expressedby the Weber-Fechnerlaw,ΔL ≈ 0.01 L –or– ΔL/L≈ 0.01where ΔL is the smallestluminance difference theeye can distinguish. (Thisequation is approximate;effective ΔL tends to belarger in dark areas ofscenes and prints due tovisual interference frombright areas.)F-stop noiseExpressing noise in relative luminance units, such as f-stops, corresponds more closely to the eye’s response than standard pixel or voltage units. Noise in f-stops is obtained by dividing the noise in pixels by the number of pixels per f-stop. (I use “f-stop” rather than “zone” or “EV” out of habit; any of them are OK.)Noise in f-stops = Noise in pixels / (d(pixel)/d(f-stop)) = 1/SNR fstF-stop SNR = SNR fst = 1/(f-stop noise) = (d(pixel)/d(f-stop)) / Noise in pixelswhere d(pixel)/d(f-stop) is the derivative of the pixel level with respect to luminance measured in f-stops (log2(luminance) ). SNR is the Signal-to-Noise Ratio.The above image illustrates how the pixel spacing between f-stops (and hence d(pixel)/d(f-stop)) decreases with decreasing brightness. This causes f-stop noise to increase with decreasing brightness, visible in the figures above.Since luminance noise (measured in f-stops) is referenced to relative scene luminance, independently of electronic processing or pixel levels, it is a universal measurement that can be used to comparedigital sensor quality when sensor RAW data is unavailable.ISO 15739 noiseThe updated ISO 15739 standard (due for release in 2012; the present standard was released in 2003) has several definitions for noise, SNR, and Dynamic Range that are closely related to f-stop noise. The relationships involve more than enough math to justify putting them in a green box.Definitions:Pxl = pixel level (same as OL = output level in the ISO standard.)σpx = noise in pixels (note: standard deviation σ is equivalent to RMS noise.)L = Illumination or exposure level. (Units will not be important.)f-stops = log2Lf-stop noise = σfst = σpx /(d(Pxl)/d(f-stops)) = σpx /(d(Pxl)/d(log2L))f-stop SNR = SNR fst = 1/σfst = (d(Pxl)/d(log2L)) / σpxWe will apply the equation, d(log b(x))/dx = 1/(x ln(b)), where ln(2) = 0.6931 = 1/1.4427.Now, from the ISO standard,Incremental gain = g I = d(Pxl)/d L (Note linear units)= d(Pxl)/d(log2L) · d(log2L) / d L = 1.442(d(Pxl)/d(log2L)) / LAppendix D of the ISO 15739 standard defines total Signal-to-Noise Ratio asSNR ISO = L g I / σpx = 1.4427 L (d(Pxl)/d(log2L)) / (L σpx) = 1.4427 (d(Pxl)/d(log2L))/σpxwhich leads toSNR ISO = 1.4427 SNR fstSNR ISO is better than SNR fst by a factor of 1.4427, or equivalently, 3.18 dB.The mathematics of noise (just a taste)Amplitude distributionIn most cases, the pixel or density variations thatcomprise noise can be modeled by the normaldistribution. This is the familiar Gaussian or “bell”curve (blue on the right) whose probability densityfunction isf(x) = exp(-(x-a)2/2σ2 ) / sqrt(2πσ2 )White noise, enlarged 2X (nearest neighbor)Blurred noise, enlarged 2X (nearest neighbor)White noise has two key properties. 1. The values of neighboring pixels are uncorrelated, i.e., independent of one-another. 2. Its spectrum is flat. The white noise spectrum (above, left) shows statistical variation and a small peak at 0.25 cycles/pixel, probably caused when the image in column (B) was resized for display.For spectral (non-white) noise, neighboring pixels are correlated and the spectrum is not flat. The spectrum and image (above, right) are the result of blurring (also called smoothing or lowpass filtering), which can result from two causes. 1. The Bayer sensor demosaicing algorithm in the RAW converter causes the noise spectrum to drop by about half at the Nyquist frequency (0.5 cycles/pixel), and 2. Noise reduction (NR) software lowpass filters noise, i.e., reduces high frequency components. NR usually operates with a threshold that prevents portions of the image near contrast boundaries from blurring. But NR comes at a price: detail with low contrast and high spatial frequencies can be lost. This causes the “plasticy”appearance sometimes visible on skin. Some people love it; I don’t. (Plastic surgeons make a lot more income than I do.) The visibility of noise depends on the noise spectrum, though the exact relationship is complex. Noise at high spatial frequencies may be invisible in small prints (low magnifications) but very damaging in large prints (large magnifications). Because of the complex nature of the relationship, Kodak has established a subjective measurement of grain (i.e., noise) called Print Grain Index (Kodak Technical Publication E-58).Sharpening and unsharp masking (USM) are the inverse of blurring. They boost portions of the spectrum and cause neighboring pixels to become negatively correlated, i.e., they exaggerate the differences between pixels, making the image look noisier. Unsharp masking is often applied with a threshold that restricts sharpening to the vicinity of contrast boundaries. This prevents noise from degrading the appearance of smooth areas like skies.When poor quality lenses are used (or the image is misfocused or shaken), the image is lowpass filtered (blurred) but the noise is not. Some sharpness loss can be recovered with sharpening or。

全球2000年Landsat陆地...

全球2000年Landsat陆地...

全球2000年Landsat陆地...Enhanced Thematic MapperGeoCover?Product Description Sheet Orthorectified Landsat Enhanced Thematic Mapper (ETM+) Compressed MosaicsMosaic Product Specifications:Spectral Bands: Three Landsat ETM+ bands, each sharpened with the panchromatic band.Band 7 (mid-infrared light) is displayed as redBand 4 (near-infrared light) is displayed as greenBand 2 (visible green light) is displayed as blueCoverage: The GeoCover Landsat mosaics are delivered in a Universal Transverse Mercator (UTM) / World Geodetic System 1984 (WGS84)projection. The mosaics extend north-south over 5 degrees of latitude, andspan east-west for the full width of the UTM zone. For mosaics below 60degrees north latitude, the width of the mosaic is the standard UTM zonewidth of 6 degrees of longitude. For mosaics above 60 degrees of latitude, theUTM zone is widened to 12 degrees, centered on the standard even-numberedUTM meridians. To insure overlap between adjacent UTM zones, eachmosaic extends for at least 50 kilometers to the east and west, and 1 kilometerto the north and south.Pixel size: 14.25 meters,Contrast Enhancement: In order to maximize the information of each mosaic, EarthSat has applied a company proprietary contrast stretch known as LOCAL(Locally Optimized Continuously Adjusted Look-up-tables) stretch. Thisstretch uses multiple, locally collected histograms, to create a radiometricallyseamless blend of contrast adjustment across areas of potentially extremecontrast ranges. The suffix “__loc” is added to the mosaic name to signify theapplication of the LOCAL stretch.Absolute Positional Accuracy: ±75 (ROSE: I am comfortable with a 50 meter RMSE, but wouldn’t want to override your V&V folks) meters RMSEr.File Naming Convention: Within each UTM zone the “partitions” extend from the equator to the north and south (in the northern and southernhemisphere respectively) in 5 degree increments. The naming convention forthe mosaics is comprised of three components, separated by hyphens; the firstelement is the hemisphere (either N or S), the second is the UTM zonenumber (1-60), the last element is the latitude of the southern edge of themosaic in the northern hemisphere and the northern edge of the mosaic in thesouthern hemisphere.For example:N-13-25_2000_loc: names a LOCAL stretched mosaicpartition in the northern hemisphere, in UTM zone 13,extending between 25 and 30 degrees north latitude.S-21-10_2000_loc names a LOCAL stretched mosaic partitionin the southern hemisphere, in UTM zone 21, extendingbetween 10 and 15 degrees south latitude.GeoCover Mosaic Image Product Delivery Format: The GeoCover Landsat image mosaics are being delivered to NASA both as uncompressed colorimagery in GeoTIFF format and as compressed color imagery in MrSID TM fileformat. The data are delivered in 24-bit color. More information on theMrSID compression format and viewing software can be found at/doc/a2916701ba1aa8114431d957.html .Non-standard UTM definition: For the southern hemisphere, the GeoTiff files contain positive zone numbers with negative northing coordinates.Source (Input) Data:Imagery:Spectral Bands: Landsat ETM+ bands 7, 4, and 2,Coverage: 5x6 degrees (south of 60 degrees North), and 5x12 degrees (north of 60 degrees North),Projection/Datum: UTM / WGS84,Pixel Size: Mixture of 14.25,Interpolation Method: Cubic Convolution,Orientation: North Up,Coverage Date: Scene dependent (nominally 2000 +/- 3 years).Control:Horizontal: Image matching to 1990 GeoCover scenes where available, otherwise Landsat-7 ephemeris was used. Vertical: DTM with 3-arc second postings, where available. Where 3-arc second data not available, GTOPO30 (30-arc second ) digital elevationmodels are used.Digital Image Processing:Mosaicing:Radiometrically balanced across automatically collected seam lines.Image Enhancements:The data are spatially and spectrally unenhanced.Earth Satellite Corporation July, 2004 /doc/a2916701ba1aa8114431d957.html。

ULTIMATE IMAGING RESULT - HIGH GRADE sCMOS CAMERAS

ULTIMATE IMAGING RESULT - HIGH GRADE sCMOS CAMERAS

MOTICAM S SERIES HIGH GRADE sCMOS CAMERAS FOR AN ULTIMATE IMAGING RESULTMotics new sCMOS cameras, designed by German engineers a nd ma nufa ctured by professiona ls represent a n impressive impetus to microscope camera performance. With a member of this pro-duct line the door for new application fields opens wide. The a ccelera ted da ta rea dout ena bles to work on fast phenomena in biology and industry.FPSFAST FRAME RATESS CMOS sCMOS SENSORSNEW MOTICAM S SERIESOur well-known Motic Ima ges softwa re, a lwa ys with a cost-free upgrade option is included in all camera packages and enables to view images at ma ximum qua lity, ca pture, edit, mea sure a nd report the result... all these features are ready to work on your PC or laptop.USB 3.1USB 3.1 INTERFACEMIP 3.0 SOFTWAREMAXIMUM DETAILSIf it is about maximum resolution for display or printing, Moticam S12 is your choice. A 25 fps* fra me ra te ma na ges the la rgest da ta volume of a ll Motica m S models. If thetiniest detail is in your focus, Moticam S12 is your solution.LOW LIGHT , FAST PHENOMENAMoticam Pro S5 Lite delivers a 5MP mid-size resolution for all biological/medical fields, showing a premium sensitivity for lowlight situa tions in Fluorescence a nd Pola riza tion. The Global shutter data readout helps to accelerate data transfer from fast phenomena, from single-cell organisms to industrial work with clockworks or extrusion processes.AN UL TIMATE SOLUTION, STILL AFFORDABLEMoticam Pro S5 Plus carries a large 2/3” sensor with 5MP resolution at a maximized frame rate of 70 fps* in full resolution. This camera covers all important aspects of imaging: resolution for detail information, speed for fast phenomena and a large sensor to coverextended areas in a single shot.1/1.7"USB 3.112.0 MP2/3"USB 3.15.0 MP2/3"USB 3.15.0 MP*Frames per second under optimal lighting conditions and in compliance with computer technical requirements.MOTIC IMAGES PLUS 3.0 SOFTWARE*Frames per second under optimal lighting conditions and in compliance with computer technical requirements.***********************************1-877-977-4717 (604) 。

  1. 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
  2. 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
  3. 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。

Mosaicing of acoustic camera imagesK.Kim,N.Neretti and N.IntratorAbstract:An algorithm for image registration and mosaicing on underwater sonar imagesequences characterised by a high noise level,inhomogeneous illumination and low frame rate ispresented.Imaging geometry of acoustic cameras is significantly different from that of pinholecameras.For a planar surface viewed through a pinhole camera undergoing translational androtational motion,registration can be obtained via a projective transformation.For an acousticcamera,it is shown that,under the same conditions,an affine transformation is a goodapproximation.A novel image fusion method,which maximises the signal-to-noise ratio of themosaic image is proposed.The full procedure includes illumination correction,feature basedtransformation estimation,and image fusion for mosaicing.1IntroductionThe acquisition of underwater images is performed in noisy environments with low visibility.For optical images in those environments,often natural light is not available, and even if artificial light is applied,the visible range is limited.For this reason,sonar systems are widely used to obtain images of seabed or other underwater objects.An acoustic camera is a novel device that can produce a real time underwater image sequence.Detailed imaging methods of acoustic cameras can be found in[1].Acoustic cameras provide extremely high resolution(for a sonar)and rapid refresh rates[1].Despite those merits of acoustic cameras over other sonar systems,it still has shortcomings compared to normal optical cameras:(i)Limitation of sight range:Unlike optical cameras which have a2-D array of photosensors,acoustic cameras have a 1-D transducer array.2-D representation is obtained from the temporal sequence of the transducer array.For this reason,it can collect information from a limited range. (ii)Low signal-to-noise ratio(SNR):The size of the transducers is comparable to the wavelength of ultrasonic waves,so the intensity of a pixel depends not only on the amplitude,but also on the phase difference of the reflected signal.This is the reason for the Rician distribution of the ultrasound image noise.In addition,there is often a background ultrasound noise in underwater environments. It follows that the SNR is significantly lower than in optical images.(iii)Low resolution with respect to optical images:owing to the limitation in the transducer size,the number of transducers that can be packed in an array is physically restricted,and so is the number of pixels in the horizontal axis.For example,a mine reacquisition and identification sonar(MIRIS)has64transducers[1].(iv)Inhomogeneous insonification:The unique geometry of an acoustic camera requires the sonar device to be aligned parallel to the surface of interest,so that the whole surface falls within the verticalfield of view of the acoustic camera [1].This alignment is not always trivial,and the misalign-ment often makes dark areas in acoustic camera images. The above limitations can be addressed by image mosai-cing,which is broadly used to build a wider view image [2–4],or to estimate the motion of a vehicle[5,6].For ordinary images,mosaicing is also used for image enhancement such as denoising,deblurring,or super-resolution[7,8].There has been extensive research on image mosaicing, and its applications[9–13].However,standard methods for image registration[14,15]are not directly applicable to acoustic camera images,because of the discrepancy of image quality,inhomogeneous insonification profile,and different geometry.Marks et al.have described a mosaicing algorithm of the oceanfloor taken with an optical camera [2].Rzhanov et al.have also described a mosaicing algorithm of underwater optical images resulting in high resolution seabed maps[3].Both of them deal with a similar problem of illumination,but use different methods:image matching by edge detection and Fourier based matching, which are not directly related to our work.In addition,since their mosaicing algorithms are not intended for image quality enhancement,we need to come up with a different mosaicing algorithm.In this paper,we describe a mosaicing algorithm for a sequence of acoustic camera images.We show that an affine transformation is appropriate for images taken from an acoustic camera undergoing translational and rotational motion.We propose a method to register acoustic camera images from a video sequence using a feature matching algorithm.Based on the parameters of image registration,a mosaic image is built.During the mosaicing,the image quality is enhanced in terms of SNR and resolution.2Properties of acoustic camera imagesSonar image acquisition includes several steps,insonifica-tion,scattering,and detection of the returning signal.In this Section,we describe physical aspects of images acquired from acoustic lens sonar systems,or acoustic cameras.q IEE,2005IEE Proceedings online no.20045015doi:10.1049/ip-rsn:20045015The authors are with the Institute for Brain and Neural Systems,Brown University,Box1843Providence RI02912,USAE-mail:kio@Paperfirst received21st May2004and in revised form22nd April2005The emission and reception of ultrasound pulses by anacoustic camera is restricted within the verticalfield ofview,which isÆ5 from the plane of image acquisition. When the object is out of this verticalfield of view,itappears dark in the image as it is poorly insonified.Thisproperty makes a typical insonification pattern in acoustic camera images,which consequently brings out the necessityof insonification correction for registration and mosaicing ofthe images.The pixel size and angular coverage of acoustic camerasvary depending on the type of the camera.We have used adual-frequency identification sonar(DIDSON)system,which has been developed for the purpose of underwater target exploration[1,16].The DIDSON system has96transducers and thehorizontalfield of view is28 :A set of acoustic lenses focuses the returning signal such that each sensor has areceiving beamwidth0:3 in the horizontal axis,and10in the vertical axis.Each transducer produces an intensity profile that corresponds to a specific angle where the rangeinformation is obtained from the focal length of theacoustic lens array.The result is either a96Â512or a48Â512polar coordinates image,which has to be mapped to Cartesian coordinates in order to recover theoriginal geometry.Since the shortest range a DIDSON system can scan is 0.75m and the maximum window length is36m,the ratio ofthe largest pixel size and the smallest pixel size can be up toð36:75=0:75Þ¼49:This means,a pixel in the polar coordinates image can occupy from one to49pixels in theCartesian coordinates image.Like other B-scan ultrasonic devices,acoustic cameras obtain pixel values by calculating the intensity of the returning signal.Owing to the diverse sources of back-ground noise,the actual water pressure observed at a transducer is the sum of multiple waves with different phase.This is often approached through a random walk problem in the phase space,and brings a different noise structure called Rayleigh distribution when a signal is not present in the image,and in general the noise is modelled by Rician distribution[17].A Rician probability density function(PDF)is in many cases approximated by a Gaussian PDF with the justification that when the SNR is high,their probability density functions are almost the same[18].3Imaging geometryThe transformation between two acoustic camera images can be calculated by putting one image into the coordinate system where the image is on the xy-plane with the positive y-axis along the centre line of the image and the centre of the arc at the origin(Fig.1).During the imaging process,a point denoted by a position vector x¼ðx;y;zÞ>is projected to the polar coordinates(r,a)as followsr¼j x jð1Þa¼sinÀ1xr xyð2Þwhere r xy ðx2þy2Þ1=2;or to the Cartesian coordinates (u,v)u¼r sin a¼xffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þtan2bqð3Þv¼r cos a¼yffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þtan2bqð4Þwhere b is the angle between x and the imaging plane.Whenthe camera is translated by d x¼ðd x;d y;d zÞ>and rotated byðf;y;cÞ;the new coordinates of x arex0¼ðx0;y0;z0Þ>¼R fycðxÀd xÞð5Þwhere the rotation matrix R fyc is a3Â3matrixR fyc¼R11R12R13R21R22R23R31R32R33@1Að6ÞR11¼cos f cos cÀsin f sin y sin cR12¼Àsin f cos yR13¼cos f sin cÀsin f sin y cos cR21¼sin f cos cþcos f sin y sin cR22¼cos f cos yR23¼sin f sin cÀcos f sin y cos cR31¼Àcos y sin cR32¼sin yR33¼cos y cos c:The linear transformation T between two images shouldsatisfyu0v01B@1C A¼x0ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þtan2b0py0ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þtan2b0p1B@1C A¼Txffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þtan2bpyffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þtan2bp1B@1C A¼Tuv1B@1C Að7Þwhereb0¼tanÀ1z0ðx02þy02Þ1=2When the reflecting points of the target object are locatedroughly on a plane such as the seafloor,z can beapproximatedbyFig.1Imaging geometry of an acoustic cameraThe camera is located at the origin of the xyz-coordinate system with thepitch,yaw,and roll each set at0.In the next frameðx0y0z0-coordinateÞ;thecamera is displaced by d x¼ðd x;d y;d zÞ>and rotated byðf;y;cÞz ¼ax þby þz 0ð8Þu 0and v 0can then be rewritten asu 0¼1þz02x 02þy 021over 2ÂfðR 11þR 13a Þx þðR 12þR 13b ÞyÀðR 11d x þR 12d y þR 13ðd z Àz 0ÞÞg v 0¼1þz 02x þy 12ÂfðR 21þR 23a Þx þðR 22þR 23b Þy ÀðR 21d x þR 22d y þR 23ðd z Àz 0ÞÞgð9Þwhere a ,b ,and z 0=ðx 02þy 02Þ1=2are sufficiently small thattheir squares are negligible.For example,the maximum deviation angle b max of a DIDSON system is 5 ;thus tan 2b max ¼0:0038:Up to a first order of approximation,we haveu0v 010@1A ¼T u u 10@1A ð10ÞwhereT ¼R 11þR 13a R 12þR 13b ÀðR 11d x þR 12d y þR 13ðd z Àz 0ÞÞR 21þR 23a R 22þR 23b ÀðR 21d x þR 22d y þR 23ðd z Àz 0ÞÞ010B @1C Að11ÞThis serves as a first order approximation of the transform-ation between two acoustic camera images.Further approximation will be studied in subsequent work by segmenting the image into local planes depending on levels of elevation.The six unknown parameters of the affine transformation can be obtained by matching features in two images.However,other parameters such as R ij ;a ,b ,or d x in (11)cannot be figured out separately because those parameters are coupled and under-constrained.Consequently,under the above approximation,it is impossible to reconstruct the precise motion of the acoustic camera merely based on image registration parameters.4MethodologyThe typical four steps of image registration are:featuredetection,feature matching,transformation estimation,and image resampling and transformation [14].Feature detection is the process of finding objects such as corners,edges,line intersections,etc.,manually or automatically.The features from the sensed image are paired with the corresponding features in the reference image in the second step.In the third step,the transformation is estimated based on the displacement vector of each feature.Once the mapping between images is established,the multiple images are combined to generate a mosaic image.In our work,we have found that high curvature points can be useful as features of interest in acoustic camera images.The sum of squared difference is used to measure the dissimilarity between two images in the second step.Transformation parameters are estimated via a random sampling based method.After the parameters of the affinetransformation are obtained,all images are combined by weighted average.4.1Coordinate mapping and inhomogene-ous insonification equalisationIn order to restore the spatial homogeneity of the image,a transformation to the Cartesian coordinates has to be performed.Owing to the fact that the field of view in the angular coordinate of different sensors does not overlap,the resulting pixel size in the Cartesian coordinates is not homogeneous.Therefore,nearest neighbour interpolation was applied to fill the gaps in the image in the Cartesian coordinate system.Owing to the acoustic acquisition of images,which was performed by insonifying the area with a single source,an inhomogeneous intensity profile is obtained.This has to be corrected for efficient image registration and mosaicing.For example,Rzhanov et al.have subtracted a 2-D polynomial spline of the image from the original image [3].Previous work on separation of illumination from reflectance was based on the Retinex theory [19];The Retinex theory was designed for optical images with low ing a homomorphic filtering method with a Gaussian retinex surround [20],Jobson et al.estimated the illumination of an image,and reconstructed the image under uniform illumination.While noise is stronger with an acoustic camera,we demonstrate that,when including the noise term in the model,the sum of squared difference is still a good dissimilarity measure after the retinex rendition.The noisy image is modeled byI ðu Þ¼L ðu Þ^I ðu Þþ s Gðu Þð12Þwhere I (u )is the observed image,L (u )the insonificationintensity,^Iðu Þthe normalised image under uniform insonification,and s G ðu Þa Gaussian noise with standarddeviation s G at u .The estimated insonification intensity ~Lis calculated by applying a Gaussian filter to the originalimage,~L ðu Þ¼I ðu Þ eÀj u j 2=2s 2and the estimated uniform insonification image is~Iðu Þ¼L ðu Þ~L ðu Þ^I ðu Þþ ðu Þ~L ðu Þ’^Iðu Þþ ðu Þ~Lðu Þð13ÞThe sum of squared difference between two uniform insonification images isSSD 1;2¼Z Zð~I 1ðu ÞÀ~I 2ðu ÞÞ2d 2u ’Z Zð^I1ðu ÞÀ^I 2ðu ÞÞ2d 2u þZ Z 1ðu Þ~L1ðu ÞÀ 2ðu Þ~L 2ðu Þ2d 2u ð14ÞThe second integral in (14)is independent of the true image,and may be regarded as a constant,provided the noise is uniform.A regularisation factor that is added to ~Lprevents erroneously excessive intensity in the equalised image from the speckles in low insonification regions.The computation speed is improved by calculating the convolu-tion in the frequency domain.4.2Feature detection and putative matching Feature detection and matching are computationally demanding.A Gaussian pyramid algorithm has been proposed as a multiscale approach for efficient feature detection and matching[21,14].As mentioned in Section 2,a pixel in the polar coordinates image corresponds to1 or several pixels in the mapped image.For example,in an image with the range8.25–44.25m,the number of pixels that correspond to a single pixel in the polar coordinates image varied from1to28.Magnified pixels result in jagged edges in the mapped image.In our images,feature detection at the third level of the Gaussian pyramid reduces false detection of corners at the jagged edges.Feature detection and putative matching is initialised by translational displacement detection.Translational displa-cement between the sensed image and the reference image is calculated by an exhaustive search on the fourth level of the Gaussian pyramid.This process drastically reduces the area of exhaustive search.After translation is estimated,high curvature points of the sensed image are detected using the Harris corner detector [22].The second moment matrix M is computed using the following relationshipM¼eÀx T x=2s2s ððH IÞðH IÞTÞð15Þwhere s s is the scale factor of the corner,and H I is the gradient vector of the image.The response after the Harris corner detection isR¼det MÀk T rðMÞ2ð16Þwhere k is set to0.04.The local maxima of R correspond to corners.These corners are matched to the corresponding points in the reference image by another exhaustive search on the third level of the Gaussian pyramid.4.3Transformation estimationImage changes due to the sonar system movement are modelled by an affine transformation as derived in the previous Section.The affine transformation describes the image changes by yaw,small pitch and roll and translational movement of the sonar system.This is valid when multiple objects are not present at the same range and angle.This is the case with the great majority of images in our dataset[1].The detailed procedure of the algorithm is as follows: (i)Feature points estimation:Using the Harris corner detector,compute50interest points from an equalised acoustic camera image.(ii)Corresponding points search:For a square patch around each feature point in the sensed image,find the sub-pixel-wise displacement in the next image,using a cross-correlation based matching.(iii)Transformation parameter estimation:Repeat the following(1)–(3)for1000samples.(1)Select3putative matching pairs.(2)Using the matching pairs,estimate the parameters of the affine transform.(3)Find the inliers of the estimated transform,and repeat (2)with the inliers until the estimated inliers are stabilised.(iv)Set a certain k percentile to define a threshold n of feature points.Then,find the n pairs of points that are closest to each other.The least mean squared error of the pairs is used as the criterion.In general,we can get better registration if wefind and match more feature points from images.However,the structure in an acoustic camera image is not sophisticated owing to the resolution and noise.In step(i),50turned out to be a reasonable number of feature points that we can reliablyfind from most of acoustic camera images.We use the criterion of least square error of k%of samples,where k is determined empirically.It is similar to the least-median of squares(LMS)method[23]in addition to the random sample consensus(RANSAC) algorithm[24],but it differs in that it can have a lower breakdown point(k instead of0.5of LMS),and it uses the mean squared error instead of the k percentile as the measure of error.It works well with a small number of feature point pairs with a high percentage of outliers. In addition,it yields a measure of goodness of the transformation,which helps to decide whether to continue mosaicing or to stop,for example,because the risk of mismatch ishigh.Fig.2Original and transformed images,and estimated and corrected insonification imagesa An original polar coordinates image from a DIDSON system.The resolution is48Â512:The range coverage is8.25m to44.25m and the angle coverage is28b Panel a mapped to the Cartesian coordinates.The resolution is512Â844c The estimated insonification of panel b.This image was produced by convolving panel b with a2-D isotropic Gaussian kernel with s G¼50 pixels.The regularisation constant was set to15d The estimated uniform insonification image of bProvided that there are about 40%of inlier feature point pairs,the probability that three inlier pairs are drawn is 0.48with 10samples,0.9987with 100samples,and 1À10À27with 1000samples.The repetition time in step (iii)may vary depending on the quality of the images.4.4Mosaicing and resolution enhancement via image fusionAfter the registration,a mosaic image is constructed.Since the noise is present regardless of the insonification condition,it can deteriorate the mosaic image if not treated properly.For example,if we average well-insonified images and poorly-insonified images,the SNR will be deteriorated because noise may accumulate.In this case,mosaicing via averaging can be described as the following relationshipI mosaic ðu Þ¼1N XN i ¼1I i ðT iu Þð17Þwhere T i is the transformation matrix from the perspective of the mosaic image to the perspective of the i th image.The SNR of the mosaic imageisFig.3Matched (circle)and non-matched (triangle)feature points obtained from the third level of the Gaussian pyramid using the Harris corner detectionFeatures from a sensed image are paired with corresponding points in the reference image.A 15Â15patch around each feature point in the sensed image is matched with the same sized patch from the corresponding 21Â21area in the reference image.The outliers (features with weaker matching)are defined by those pairs with higher matching error after the estimatedtransformationFig.4Demonstration of weighted averaging effectMosaicing was performed with 38images which were averaged after the corresponding motion compensation transformation was applied to each of them a Uniform average of the whole imagesb Weighted average,in which insonification profile was utilised during averaging (see Section 4.4for details)c Same target from different image sequenced Same target from different image sequence utilising the insonification profile during averagingSNR mosaic ðu Þ¼PiL i ðT i u Þsð18Þwhere L i ðu Þis the insonification intensity of the i th image at u .Note that the SNR is a function of u because the insonification intensity varies within the image.In our algorithm,poorly insonified regions receive lower weight in the averaging.Denote the weight of the i th image by a i ðu Þ;where P i a i ðT i u Þ¼1:Then,the mosaic image isI mosaic ðu Þ¼Xa i ðT i u ÞI i ðT i u Þð19Þof which the SNR mosaic isSNR mosaic ðu ;a 1;...;a N Þ¼P a i L i ðÞs ffiffiffiffiffiffiffiffiffiffiffiPa 2ip ffiffiffiffiffiffiffiffiffiffiffiP L 2i p s ð20ÞEquality holds when a k ðu Þ¼L k ðu Þ=P L i ðT À1k T i u Þ:Thus,the maximum SNR of the mosaic image is achieved when the transformed images are combined as followsI mosaic ðu Þ¼PL i ðT i u ÞI i ðT i u ÞP L i ðT i u Þð21ÞThis weighted averaging method reduces the influence of noise in poorly insonified regions.5ResultsThe algorithm was tested on a boat wreckage sequence [Note 1].A DIDSON system scanned a shipwreckatFig.5Resolution enhancement by averaging imagesa Ship wreckage image juxtaposed with a mosaic image of 5consecutive frames followed by a geometric transformation (see Section 4.3)b Coral image with a mosaic image of 7consecutive framesNote 1:The data has been provided by E.O.Belcher from Applied Physics Laboratory,University of Washington under ONR support.approximately30metres depth for285seconds and took 446frames of images.About40frames among them show the vessel from head to stern,and another40frames show it from stern to head.The body of the ship exposed in each frame is less than20%in each frame.The algorithm was applied to those two sub-sequences to build two mosaic images.Figures2a and b depict the same acoustic image in the original polar coordinates and the transformed Cartesian coordinates,respectively.A collection of pixels with the same pixel intensity can be seen in the Cartesian coordinates image.Estimated insonification based on the method described in Section4.1is depicted in Fig.2c.The insonification corrected image is depicted in Fig.2d.The insonification correction,which equalises the image,increases the dynamic range of the averaged(mosaiced)image.Figure3depicts two consecutive acoustic images together with a set of matched(circle)and non-matched (triangle)feature points.These matched feature points in the reference image,which were found using the cross-correlation of patches around the feature points in the sensed image,are used to estimate the geometric trans-formation between the two images.Cross-correlation was found to be more robust than a conventional approach[25]in which features are independently found and matched between the two images.This is a consequence of the high noise in the image and the fact that the exact location of the features is not well defined.Figure4represents the main result of the paper,a mosaic image of multiple acoustic images.The mosaiced image contains information which spans mul-tiple frames,each frame corresponding to a small portion of the insonified object.The combination images,which have been transformed to be in the same coordinate system,provide subpixel image resolution enhancement. Left panels of Figs.5a and b show the original single frames detail of the target before mosaicing.The resolution enhancement follows from the fact that one pixel in the original polar coordinate system is mapped to multiple pixels with the same intensity in the Cartesian coordinate system.Different frames lead to partial overlap of these multiple pixels,so that after averaging, a subpixel resolution is achieved(see right panels of Figs. 5a and b).Averaging of different acoustic images after bringing them to the same coordinate system(same viewpoint)leads to the classical effect of denoising.This is clearly seen in Fig.4on the whole target,and in particular in the comparison of two frames of the targets in Fig.5.The top two panels of Fig.4depict a mosaiced image from the same sequence of acoustic images.In panel b,the insonification profile was utilised during averaging.Panels c and d represent the same target from a different acoustic image sequence with panel d utilising the insonification profile during averaging.6ConclusionAcoustic camera technology is becoming essential for underwater exploration in noisy environments with low visibility.The acoustic camera,with its specific sensor design,poses some challenges in terms of image resolution, noise removal and area coverage.In this paper,we have presented a complete algorithm to achieve image mosaicing,denoising and resolution enhancement from a sequence of acoustic camera images.We described the steps that were required to achieve this mosaicing.This included modelling the specific geometry of acoustic camera images which sharply differs from pinhole camera geometry.The different geometry,and in particular,the fact that the images are acquired in a polar coordinate system, complicates the search and matching of feature points in consecutive images.Moreover,in this particular geometry, pixels in the polar coordinate system are mapped to a collection of pixels with the same intensity in the Cartesian coordinate system.Since consecutive images were taken from different viewpoints,a subpixel enhancement effect was achieved in the process of averaging in addition to the denoising effect.We have presented a novel method in which features extracted by the Harris corner detector are matched locally to the reference image via cross-corre-lation.This method was found to be more robust than a conventional approach in which features are found inde-pendently and matched between two images.In particular, this is more pronounced when the number of pixels available for feature comparison is limited.7AcknowledgmentsThis work was partly supported by ONR grant N00014-02-C-0296.The authors thank E.O.Belcher for providing full details about the data.Leon N.Cooper and other members of IBNS have provided valuable comments.8References1Belcher,E.O.,Matsuyama,B.,and Trimble,G.M.:‘Object identifi-cation with acoustic lenses’.Proc.Oceans’01MTS/IEEE,2001, pp.6–112Marks,R.L.,Rock,S.M.,and Lee,M.J.:‘Real-time video mosaicking of the oceanfloor’,IEEE J.Ocean.Eng.,1995,20,(3),pp.229–241 3Rzhanov,Y.,Linnet,L.M.,and Forbes,R.:‘Underwater video mosaicing for seabed mapping’.Int.Conf.Image Process.,2000,2,pp.224–227 4Castellani,U.,Fusiello,A.,and Murino,V.:‘Registration of multiple acoustic range views for underwater scene reconstruction’,Comput.Vis.Image Underst.,2002,87,pp.78–895Garcı´a,R.,Cufı´,X.,and Pacheco,L.:‘Image mosaicking for estimating the motion of an underwater vehicle’.5th IFAC Conf.on Manoeuvreing and Control of Marine Craft,20006Negahdaripour,S.,Xu,X.,and Khamene,A.:‘A vision system for real-time positioning,navigation and video mosaicing of seafloor imagery in the application of rovs/auvs’.IEEE Workshop on put.Vis.,1998,pp.248–2497Capel,D.,and Zisserman,A.:‘Computer vision applied to super-resolution’,IEEE Signal Process.Mag.,2003,20,(3),pp.72–868Smolic,A.,and Wiegand,T.:‘High-resolution video mosaicing’.Int.Conf.on Image Process.,2001,3,pp.872–8759Hansen,M.,Anandan,P.,Dana,K.,Wal,G.,and Burt,P.:‘Real-time scene stabilization and mosaic construction’.Proc.IEEE Workshop put.Vis.,1994,pp.54–6210Mann,S.,and Picard,R.W.:‘Virtual bellows:Constructing high quality stills from video’.Proc.IEEE Int.Conf.Image Process.,1994, pp.363–36711Irani,M.,Anandan,P.,Bergen,J.,Kumar,R.,and Hsu,S.:‘Mosaic representations of video sequences and their applications’,Signal Process.,Image Commun.,1996,8,(4),pp.327–35112Irani,M.,Hsu,S.,and Anandan,P.:‘Video compression using mosaic representations’,Signal Process.,Image Commun.,1995,7, pp.529–55213Sawhney,H.S.,and Ayer,S.:‘Compact representation of videos through dominant multiple motion estimation’,IEEE Trans.Pattern Anal.Mach.Intell.,1996,18,(8),pp.814–83014Zitova´,B.,and Flusser,J.:‘Image registration methods:a survey’, Image put.,2003,21,pp.977–100015Brown,L.G.:‘A survey of image registration techniques’,ACM Comput.Surv.,1992,24,(4),pp.325–37616Belcher,E.O.,Dinh,H.Q.,Lynn,D.C.,and Laughlin,T.J.:‘Beamform-ing and imaging with acoustic lenses in small,high-frequency sonars’.Proc.Oceans’99MTS/IEEE,1999,pp.1495–149917Wagner,R.F.,Smith,S.W.,Sandrik,J.M.,and Lopez,H.:‘Statistics of speckle in ultrasound b-scans’,IEEE Trans.Sonics Ultrason.,1983,30,(3),pp.156–16318Sijbers,J.,den Dekker,J.,Scheunders,P.,and Van Dyck, D.:‘Maximum-likelihood estimation of rician distribution parameters’, IEEE Trans.Med.Imaging,1998,18,(3),pp.357–361。

相关文档
最新文档