武汉理工大学数字图像处理英文课件 Chapter2（3）.pptVIP

Making Ellipses Fitting Ellipses (1/3) Fitting Ellipses (2/3) Fitting Ellipses (3/3) Matlab Code Example * function img = ellipse(sz, cen, rad, theta, thres) x = 1:sz(2); y = 1:sz(1); x = ones(sz(1),1)*x; y = y*ones(1,sz(2)); theta = theta*pi/180; p = cos(theta)*(x-cen(1)) - sin(theta)*(y-cen(2)); q = sin(theta)*(x-cen(1)) + cos(theta)*(y-cen(2)); r = sqrt((p/rad(1)).^2+(q/rad(2)).^2); img = rthres; im = ellipse([200 200],[100 100],[80 10],20,1); imshow(im); imwrite(im,ellipse.tif); The centroid of the ellipse is calculated from for a set of N data points. Next we define a rotated set of axes (u, v) given by We then find the value of q that minimizes the sum of deviations We can rewrite Svv as where We can simplify Svv by using double angle formulas: The desired value for q is found from which has the solution For this value of q the variance along the major and minor axis is proportional to The major and minor radii of an ellipse may be found from Major radius Minor radius function [cen, elips] = find_ellipse(mask) img = im2double(mask); sz = size(img); x = 1:sz(2); y = 1:sz(1); x = ones(sz(1),1)*x; y = y*ones(1,sz(2)); tot=sum(sum(img)); cen(2) = sum(sum(x.*img))/tot; cen(1) = sum(sum(y.*img))/tot; angle = atan2(ky,kx)/2; kx = cos(angle); ky = sin(angle); sr = sxx+syy; elips(1) = sqrt(2*(sr+sd)); elips(2) = sqrt(2*(sr-sd)); elips(3) = angle*180/pi; x = x - cen(2); y = y - cen(1); sxx = sum(sum(x.*x.*img))/tot; sxy = sum(sum(x.*y.*img))/tot; syy = sum(sum(y.*y.*img))/tot; kx = sxx - syy; ky = 2*sxy; sd = sqrt(kx.^2+ky.^2); fprintf(orientation angle: %g deg\n,-elips(3)); fprintf(major %g minor %g\n,elips(1),elips(2)); *