mubanpipei.rar_互相关法_子图匹配

function [x, y]=SSDA_Match(img, subimg) %-------------------------------------------------------------------------% %该函数用来查找原始图像img中的子区块subimg的位置，位置值取匹配区域的左上角坐标。 %函数使用互相关法实现。为了减小搜索区域，先将原图像和子图像抽取至较小范围，获取 %相应位置后，再在原图像相应点附近详细搜索。 %返回值x、y是相应的坐标值。 %-------------------------------------------------------------------------% [M, N] = size(img); %获取原图像、子图像大小 [Msub, Nsub] = size(subimg); Mtmp = Msub; Ntmp = Nsub; img = double(img); subimg = double(subimg); if (M > Msub) & (N > Nsub) times = 0; while (Mtmp > 32) | (Ntmp > 32) %获取可压缩的次数times Mtmp = Mtmp/2 ; Ntmp = Ntmp/2 ; times = times + 1; end img1 = img(1:2.^times:M, 1:2.^times:N); %得到压缩图像 subimg1 = subimg(1:2.^times:Msub, 1:2.^times:Nsub); [Ms, Ns] = size(img1); [Mssub, Nssub] = size(subimg1); CorRes = zeros(Ms-Mssub+1, Ns-Nssub+1); %创建并且计算压缩图互相关结果 for delta_i = 1:Ms-Mssub+1 for delta_j = 1:Ns-Nssub+1 tmp1 = sum(sum(img1(delta_i:delta_i+Mssub-1, delta_j:delta_j+Nssub-1).*subimg1(1:Mssub, 1:Nssub))); tmp2 = sqrt(sum(sum(img1(delta_i:delta_i+Mssub-1, delta_j:delta_j+Nssub-1).^2))); tmp3 = sqrt(sum(sum(subimg1(1:Mssub, 1:Nssub).^2))); CorRes(delta_i, delta_j) = tmp1/(tmp2*tmp3); end end [tmp, k] = max(CorRes); [tmp, kk] = max(tmp); xs = k(kk); ys = kk; %取得最大互相关坐标 x1 = 1 + (xs-1) .* (2.^times); %找到原始图像带搜索的区域 x2 = min([xs .* (2.^times) M]); %防止坐标越界 y1 = 1 + (ys-1) .* (2.^times); y2 = min([ys .* (2.^times) N]); CorRes1 = zeros(x2-x1+1, y2-y1+1); %创建并且计算原图互相关结果 for delta_i = x1:x2 for delta_j = y1:y2 tmp1 = sum(sum(img(delta_i:delta_i+Msub-1, delta_j:delta_j+Nsub-1).*subimg(1:Msub, 1:Nsub))); tmp2 = sqrt(sum(sum(img(delta_i:delta_i+Msub-1, delta_j:delta_j+Nsub-1).^2))); tmp3 = sqrt(sum(sum(subimg(1:Msub, 1:Nsub).^2))); CorRes1(delta_i-x1+1, delta_j-y1+1) = tmp1/(tmp2*tmp3); end end [tmp, k] = max(CorRes1); [tmp, kk] = max(tmp); delta_x = k(kk) - 1; delta_y = kk - 1; %取得最大互相关坐标 x = x1 + delta_x; %得到精确坐标 y = y1 + delta_y; % str1 = strcat('计算得到的精确匹配坐标为 [',num2str(x),','num2str(y),']') else error('子图的维数比待查图像大'); %子图大小越界 end