Digital-Image-Processing-4教学文案

D i g i t a l-I m a g e-P r o c e s s i n g-4Digital Image ProcessingComputer Project Report IV Contrast Enhancement, Histogram Equalization, Spatial Filtering, & Edge DetectionStudent ID: 20091613310032Name: Xiaopeng JiDate: April 13, 2012A. Objectivesi. Study the effects of contrast enhancement histogram equalization andhistogram stretching.ii. Study the effects of histogram equalization via HISTEQ function.iii. Utilizing the SUBPLOT function to display images.iv. Study the effects of spatial filtering using IMNOISE, FSPECIAL, and MEDFILT2 functions.v. Study image noise and how the median and average filters remove noise.vi. Study different methods of edge detection for use on noisy images specifically motion blur, Gaussian noise and filtered Gaussian noise via averaging.vii. Formulate real-world applications which utilize the above functions and concepts.B. MethodsThis project will utilize the functions of MATLAB Image Processing Toolbox to perform Contrast Enhancement, Histogram Equalization, Spatial Filtering using averaging and median filters, and an Edge detection algorithm experiment.1. IMRESIZE function utilization for image expansion.2. Contrast stretching using IMADJUST and STRETCHLIM functions.3. Histogram equalization using HISTEQ function.4. Data display utilizing SUBPLOT function.5. Spatial filtering using IMNOISE, FSPECIAL, and MEDFILT2.6. Edge detection on noisy images using Matlab noise and edge detection functions.C. Results1. Load image saltAndPepper.gif and use the IMRESIZE function tointer polate by a factor of 4 using ‘bi-cubic’ interpolation.At the command prompt enter :>>saltandpepper_filename='saltAndPepper.gif';saltandpepper_image=imread(saltandpepper_filename);picture_saltandpepper_scale=imresize(saltandpepper_ima ge,[60*4 59*4],'bicubic');imwrite(picture_saltandpepper_scale,'saltandpepper_bic ubic.jpg', 'jpg');The original image, saltAndPepper.gif, was spatially expanded 4x in each dimension using bi-cubic interpolation. The IMRESIZE function was utilized to perform the operation.Figure 1: Original image.Figure2: saltAndPepper.gif spatially expanded by 4x.2. Contrast Stretching – Use the MATLAB function IMADJUST to perform the contrast stretch.a) Use imadjust(Image, [LOW_IN, HIGH_IN],[]); in this format. [LOW_IN, HIGH_IN] must be in the [0,1] range so the vector must be scaled by the maximum gray level that the image is quantized to (i.e. 8-bits a 256 gray levels).At the command prompt enter :>>saltandpepper_imadjust=imadjust(picture_saltandpepper_c ale,[0 (180/255)],[]);imwrite(saltandpepper_imadjust,'saltandpepper_imadjust.jpg','jpg');Figure 3: Adjusted figure 2 such that the lowest and highest intensities are 0and 255 respectively.b) Compute the mean gray level (function: MEAN2) and the standard deviation (function: STD2) and design a mapping function for LOW_IN and HIGH_IN based on the following equations:LOW_IN=(mean-(standard deviation)/2)/256HIGH_IN=(mean+(standard deviation)/2)/256At the command prompt enter :>>saltandpepper_mean=mean2(picture_saltandpepper_scale) saltandpepper_std=std2(picture_saltandpepper_scale)low_in=(saltandpepper_mean-(saltandpepper_std)/2)/256 high_in=(saltandpepper_mean+(saltandpepper_std)/2)/256 saltandpepper_mean=imadjust(picture_saltandpepper_scal e,[low_in high_in],[]);imwrite(saltandpepper_mean,'saltandpepper_mean.jpg','j pg');Figure 2: The histogram of cameraman.tifsaltandpepper_mean = 60.8439saltandpepper_std = 30.3702low_in = 0.1784high_in = 0.2970Figure 4: IMADJUST performed on Figure 2 with LOW_IN and HIGH_IN equalto above calculationsc) Use the function STRETCHLIM to compute the tolerance for the IMADJUST function. The function STRETCHLIM will compute a vector corresponding to [LOW_IN, HIGH_IN]. Use the vector calculated by STRETCHLIM to in the IMADJUST function described in a.) and compute the image transformation.At the command prompt enter :>>low_high = stretchlim(picture_saltandpepper_scale, 0) saltandpepper_stretchlim=imadjust(picture_saltandpepper_s cale,low_high, []);imwrite(saltandpepper_stretchlim ,'saltandpepper_stretchlim.jpg', 'jpg');low_high = (00.7059)Figure 5: STRETCHLIM performed with IMADJUST on Figure 23. Histogram Equalization – Use the function HISTEQ to compute the histogram equalization of the original image.4. Data Display – Use the SUBPLOT command to display your results.At the command prompt enter:>>imhist(picture_saltandpepper_scale);saltandpepper_EqualizedImage =histeq(picture_saltandpepper_scale);imwrite(saltandpepper_EqualizedImage,'saltandpepper_Equal izedImage.jpg', 'jpg');colormap('gray');subplot(2,2,1);image(picture_saltandpepper_scale);title('Original Image');subplot(2,2,2);image(saltandpepper_mean);title('Mean and Std Image');subplot(2,2,3);image(saltandpepper_stretchlim);title('StretchLim Image');subplot(2,2,4);image(saltandpepper_EqualizedImage);title('Equalized Image');figure;subplot(2,2,1);imhist(picture_saltandpepper_scale);title('Original Image');subplot(2,2,2);imhist(saltandpepper_mean);title('Mean and Std Image');subplot(2,2,3);imhist(saltandpepper_stretchlim);title('StretchLim Image');subplot(2,2,4);imhist(saltandpepper_EqualizedImage);title('Equalized Image');Figure 3: After histogram equalization5. Spatial Filtering – Use the functions IMNOISE, FSPECIAL, and MEDFILT2 to perform spatial filtering of noise corrupted images.a) Using the histogram equalized image of 3) use the IMNOISE function to generate two noise corrupted images.noise_gaussian=imnoise(Image,’gaussian’,0,0.05);noise_saltAndpepper=imnoise(Image,’salt & pepper’, 0.02);Use the data display technique in 4) to display the original histogram enhanced image, and the two noise corrupted images.Atnoise_gaussian=imnoise(saltandpepper_EqualizedImage,'gaussian',0,0.05 );noise_saltAndpepper=imnoise(saltandpepper_EqualizedImage,'salt & pepper', 0.02);figure;colormap('gray');subplot(1,3,1);imagesc(saltandpepper_EqualizedImage);title('Equalized Image');subplot(1,3,2);imagesc(noise_gaussian);title('Noise Gaussian');subplot(1,3,3);imagesc(noise_saltAndpepper);title('Noise Salt and Pepper');b) Use the function FSPECIAL to design averaging filters of size (3x3), (5,5), and (7x7).average_filter_3x3=fspecial('average',[3 3])average_filter_5x5=fspecial('average',[5 5])average_filter_7x7=fspecial('average',[7 7])c) Using the filter masks generated in b), use the function IMFILTER to perform the two-dimensional convolution function of the mask with thenoise_gaussian image that was generated in a). Use the display technique in 4) to display the noise_gaussian image and the three averaging filtered results.average_gaussian_3x3=imfilter (noise_gaussian,average_filter_3x3); average_gaussian_5x5=imfilter (noise_gaussian,average_filter_5x5); average_gaussian_7x7=imfilter (noise_gaussian,average_filter_7x7);figure;colormap('gray');subplot(2,2,1);imagesc(noise_gaussian);title('Noise Gaussian');subplot(2,2,2);imagesc(average_gaussian_3x3);title('Average Gaussian 3x3');subplot(2,2,3);imagesc(average_gaussian_5x5);title('Average Gaussian 5x5');subplot(2,2,4);imagesc(average_gaussian_7x7);title('Average Gaussian 7x7');d) Use the MEDFILT2 function to perform median filtering on thenoise_saltAndpepper image generated in a). Design the median filters to work with window sizes of (3x3), (5x5), and (7x7). Use the display technique in 4) to display the noise_saltAndpepper image and the three median filtered results. average_saltAndpepper_3x3=medfilt2 (noise_saltAndpepper,[3 3]); average_saltAndpepper_5x5=medfilt2 (noise_saltAndpepper,[5 5]); average_saltAndpepper_7x7=medfilt2 (noise_saltAndpepper,[7 7]);figure;colormap('gray');subplot(2,2,1);imagesc(noise_saltAndpepper);title('Noise saltAndpepper');subplot(2,2,2);imagesc(average_saltAndpepper_3x3);title('Median saltAndpepper 3x3');subplot(2,2,3);imagesc(average_saltAndpepper_5x5);title('Median saltAndpepper 5x5');subplot(2,2,4);imagesc(average_saltAndpepper_7x7);title('Median saltAndpepper 7x7');6. Edge Detection – Use the functions EDGE, FSPECIAL, IMAGESC to perform and evaluate each detectors performance in several different scenarios. Load image wheel.gif .a.) Original Image: Use the EDGE function to generate masks for Roberts, Canny, Sobel, and Prewitt operators. Use the IMAGESC function and COLORMAP(GRAY) to plot the result of each edge mask. State which operator gives the best performance and why you think so.wheel_filename='wheel.gif';wheel_image=imread(wheel_filename);BW1=edge(wheel_image,'prewitt');BW2=edge(wheel_image,'canny');BW3=edge(wheel_image,'sobel');BW4=edge(wheel_image,'roberts');figure;colormap('gray');subplot(2,2,1);imagesc(BW1);title('Prewitt'); subplot(2,2,2); imagesc(BW2);title('Canny'); subplot(2,2,3); imagesc(BW3);title('Sobel'); subplot(2,2,4); imagesc(BW4);title('Roberts');b.) Motion Blur: Use the function FSPECIAL (i.e., fspecial('motion',50,45)) to generate motion blur. This will simulate blur due to camera or image platform motion. Use the same procedure as in a.) to evaluate your results.motionfilter=fspecial('motion',50,45);wheel_filter=imfilter (wheel_image,motionfilter);BW1=edge(wheel_filter,'prewitt');BW2=edge(wheel_filter,'canny');BW3=edge(wheel_filter,'sobel');BW4=edge(wheel_filter,'roberts');figure;colormap('gray');subplot(2,2,1);imagesc(BW1);title('Prewitt');subplot(2,2,2);imagesc(BW2);title('Canny');subplot(2,2,3);imagesc(BW3);title('Sobel');subplot(2,2,4);imagesc(BW4);title('Roberts');c.) Gaussian Noise: Use the function IMNOISE (i.e.,imnoise(Image,'gaussian',0,0.001);) to generate a noise corrupted image. This experiment is in similar context to b.) as it illustrates how the gradient information degrades due to system parameters such as platform motion and signal to noise ratio. Use the same procedure as in a.) to evaluate your results.gaussian_filter=imnoise (wheel_image,'gaussian',0,0.001);BW1=edge(gaussian_filter,'prewitt');BW2=edge(gaussian_filter,'canny'); BW3=edge(gaussian_filter,'sobel'); BW4=edge(gaussian_filter,'roberts');figure;colormap('gray');subplot(2,2,1);imagesc(BW1);title('Prewitt');subplot(2,2,2);imagesc(BW2);title('Canny');subplot(2,2,3);imagesc(BW3);title('Sobel');subplot(2,2,4);imagesc(BW4);title('Roberts');d.) Filtered Gaussian Noise: Use the 3 x 3 Averaging filter designed using FSPECIAL in section 5b.) to filter the noise image generated in c.). Use the same procedure as in a.) to evaluate your results.At the command prompt enter:avg_gaussian_filter=imfilter(gaussian_filter,average_filter_3x3);BW1=edge(avg_gaussian_filter,'prewitt');BW2=edge(avg_gaussian_filter,'canny');BW3=edge(avg_gaussian_filter,'sobel');BW4=edge(avg_gaussian_filter,'roberts');figure;colormap('gray');subplot(2,2,1);imagesc(BW1);title('Prewitt');subplot(2,2,2);imagesc(BW2);title('Canny');subplot(2,2,3); imagesc(BW3);title('Sobel'); subplot(2,2,4); imagesc(BW4);title('Roberts');。