武汉理工大学数字图像处理英文课件 Chapter3（5）.pptVIP

Linear Regression Method of Least Squares (1/2) Method of Least Squares (2/2) Matlab Example RGB Components NTSC Y Channel Luminance Hue and Saturation White – Color Components Saturation Inverse Saturation Image Multi-Channel Linear Regression Matlab Example Reshaping Matrices Reshaping Matrices Reshape Reshape Example * Assume that image Y is point-by-point linearly related to image X First we define some notation for the mean of an image: We can use the mean to eliminate k0 The method of least-squares is used to find k1. The total least-squared deviation is: The expression can be simplified as follows: where If the total least-squared deviation is We can minimize s2 with respect to k1 Then We can find k0 from gm1 = mean2(gray1) gm2 = mean2(gray2) gc1 = gray1 - gm1; gc2 = gray2 - gm2; sxy = mean2(gc1.*gc2); sxx = mean2(gc1.*gc1); k1 = sxy/sxx k0 = gm2 - k1*gm1 diff2 = gray2 - k0 - k1*gray1; cd = 0.5*(diff2+1); imwrite(cd, cd.jpg); sigma = std2(diff2); gm1 = 0.5563 gm2 = 0.5579 k1 = 0.9694 k0 = 0.0186 sigma = 0.0433 original Luminance YIQ Intensity HSI Value HSV Lightness HLS Luminance is 1 when R=G=B=1 Saturation Hue white color white color Saturation is the relative proportion of pure color to the total brightness. HSV HLS Assume that image Y is point-by-point linearly related to image (R,G,B) The normal equations of least squares are then given by The total least-squared deviation is: ? rgb = imread(kitchen1.tif); ? for I = 1:3 for J=1:3 C(I,J) = sum(sum(rgb(:,:,I).*rgb(:,:,J))); end end ? gray = im2double(imread(kitchen2.tif)); ? for I=1:3 d(I) = sum(sum(gray(:,:,1).*rgb(:,:,I))); end ? k = C\d; C = 1.0e+004 * 6.9239 6.4029 5.6287 6.4029 6.1035 5.3815 5.6287 5.3815 5.1229 d = 1.0e+004 * 6.4717 6.1061 5.4191 k = 0.3277 0.5617 0.1077 Sometimes it is convenient to represent an image as a vector rather than as a matrix. If X is an image in matrix form, ? X = [ 4 18 32 3; 12 11 15 5; 2 3 5 12