Digital Image Processing & Pattern Analysis (CSCE 563) Fundamentals

1. Digital Image Processing&Pattern Analysis (CSCE 563) Fundamentals Prof. Amr Goneid Department of Computer Science & Engineering The American University in Cairo

2. Image Acquisition & Display • Image Models • Sampling & Quantization • Storage Media • Image Formats • Representation in MATLAB • Loading, Saving and Display • Types of Image Operations • Types of Neighborhoods Prof. Amr Goneid, AUC

3. Real Image Model A 2-D Signal with amplitude a(x,y) representing light intensity or color at real position (x,y). Amplitude is normally a real number or an integer number. An image may contain a Region of Interest (ROI) (i.e. a sub-image) Signal y ROI a(x,y) x Prof. Amr Goneid, AUC

4. Digital Image Model An acquisition device converts a real (analog) image a(x,y) into a digital image. Sampling and quantization comprise the digitization process. The output is a matrix f(i,j) of pixels with n rows and m columns. a(x,y) Sampling & Quantization f(i,j) Prof. Amr Goneid, AUC

5. Image Acquisition:Sampling & Quantization • Sampling Sample at x,y = one pixel at f(i,j) y y x S(x,y) = average Intensity over the area x y Sampling Rate affects image size x Prof. Amr Goneid, AUC

6. Sampling: Spatial Resolution Low Resolution High Resolution Prof. Amr Goneid, AUC

7. Sampling & Quantization • Quantization Digital Image f(x,y) Fk+1 f11 f12 … f1m f21 f22 … f2m S(x,y) Fk y fij Fk-1 x Pixel value fn1 fn2 … fnm Prof. Amr Goneid, AUC

8. Quantization: Color Resolution 16 Colors 256 colors Quantization scale determines number of bits per pixel Prof. Amr Goneid, AUC

9. Image Storage Media • Magnetic & Video Tapes • Magnetic Disks • Optical Disks ( CD ) Prof. Amr Goneid, AUC

10. Storage Requirements • Image Size: n x m pixels • n , m depend on sampling rate • B = no. of bits to represent a pixel • L = no. of Quantization levels (gray levels or colors) = 2B • Storage size = n m B/8 bytes • e.g. for 128 x 128 pixels with 256 colors the size is 16 kbytes • When B = 1 we have a Binary Image (Black & White) Prof. Amr Goneid, AUC

11. Image Formats • Graphics Interchange Format ( .GIF) • Tagged Image File Format (.TIFF) • Bit Map Format (.BMP) • ISO Standard Format (.JPEG , .JPG , .JPE) • Other Formats ( .PCX , .CGI , .PNG, etc) • Viewers ( Image Viewing S/W) Prof. Amr Goneid, AUC

12. Image Formats 302 Kb (BMP) 80 Kb (jpg) Prof. Amr Goneid, AUC

13. Representation in MATLAB • Intensity Images (Gray Scale) • Indexed Images • RGB Images • Binary Images • Conversion between Formats • Loading & Saving Images • Displaying Images in MATLAB Prof. Amr Goneid, AUC

14. Intensity Images • Used for Gray Scale images • One matrix I = f(n,m) • Each pixel f(i,j) is a real number (r) with 0 =< r <= 1 ( 0 : black; 1 : white) Prof. Amr Goneid, AUC

15. Indexed Images Prof. Amr Goneid, AUC

16. Indexed Images • For color (and gray scale images) • Uses a color map and a matrix X = f(n,m) with a pixel value f(i,j) = index to the map , • map is (m x 3) array, m = no. of colors • e.g. f(i,j) = 23 G R B 23 0.3 0.1 0.2 Color Map Prof. Amr Goneid, AUC

17. RGB Images • For color images.Uses 3 matrices : R(n,m) , G(n,m) , B(n,m) • R(i,j) is the red color intensity, etc. RGB R G B Prof. Amr Goneid, AUC

18. RGB Planes A pixel at (i,j) in an RGB image has components: R(i,j) , G(i,j) , B(i,j) In MATLAB, it is possible to extract the 3 components: If F is an RGB image, FR = F(: , : , 1); FG = F(: , : , 2); FB = F(: , : , 3); B G R Summer 2008 Prof. Amr Goneid, AUC Prof. Amr Goneid, AUC 18

19. Binary Images • For Black and White images • One matrix B = f(n,m) where a pixel value is f(i,j) = 0 or 1 Prof. Amr Goneid, AUC

20. Conversion between Formats im2bw ind2rgb Indexed gray2ind rgb2ind RGB Binary gray2ind ind2gray rgb2gray im2bw edge Intensity Prof. Amr Goneid, AUC

21. Loading & Saving Images • MATLAB reads disk formats and converts to pixel matrices • e.g. for BMP files: [X,map] = imread(‘filename.bmp’) imwrite(X,map,’filename.bmp’) • Some of the types supported: BMP JPG GIF TIFF PCX • Matrix format can be saved on disk: save filenameX map • Matrix format files can be loaded from disk: load filename Prof. Amr Goneid, AUC

22. Displaying Images in MATLAB • Indexed Images: imshow(X,map) • Intensity Images: imshow(I) • RGB Images: imshow(RGB_image) • Binary Images: imshow(BW) , imshow(~BW) Prof. Amr Goneid, AUC

23. Example1input color image, output red image [X,map] = imread(‘eaglec.bmp’); % Read eaglec.bmp from disk [n,m] = size(X); % Image size save eaglec X map; % Save in matrix form _________________________________________________ load eaglec; % Load matrix form imshow(X,map); % Display image [L , C] = size(map); % L rows by 3 columns map(: , 2) = zeros(L , 1); % zero green map(: , 3) = zeros(L , 1); % zero blue % now the map contains only red imwrite( X , map , ‘eagler.bmp) % write red image Prof. Amr Goneid, AUC

24. Example2Convert to Intensity and Binary Images load eaglec; % Load matrix form imshow(X,map); % Display image I = ind2gray(X,map); % Convert to gray imshow(I); % Display BW = im2bw(I); % Convert to binary imshow(BW); % Display imshow(~BW); % Display negative Prof. Amr Goneid, AUC

25. Types of Image Operations(ref: http://www.ph.tn.tudelft.nl/Courses/FIP/noframes/fip.html ) The types of operations that can be applied to digital images to transform an input image f [m,n] into an output image f’[m,n] can be classified into three categories as shown in the Table. Prof. Amr Goneid, AUC

26. Types of Image Operations(ref: http://www.ph.tn.tudelft.nl/Courses/FIP/noframes/fip.html ) Prof. Amr Goneid, AUC

27. Neighborhood Prof. Amr Goneid, AUC

28. Types of Neighborhoods 4-connected 8-connected 6-connected Rectangular Hexagonal Prof. Amr Goneid, AUC