matlab-基于区域生长和区域合并的无监督彩色图像分割算法matlab仿真-源码

function T = PNN_fast_RGBo(Iseg,IRGB,nseg) segnum = max(Iseg(:)); n= segnum; conn = diag(ones(1,segnum)); %get boundary conn [row,col] = size(Iseg); diffx = diff(Iseg,1,2); diffy = diff(Iseg,1,1); sq1 = find(diffx); sq2 = find(diffy); for t = 1:length(sq1) v1 = Iseg(sq1(t)); v2 = Iseg(sq1(t)+ row); conn(v1,v2) = 1; conn(v2,v1) = 1; end for t = 1:length(sq2) v1 = Iseg(sq2(t)); v2 = Iseg(sq2(t)+ 1); conn(v1,v2) = 1; conn(v2,v1) = 1; end histnum = zeros(1,segnum); histval = zeros(3,segnum); for r= 1:row for c = 1:col segind = Iseg(r,c); histnum(segind) = histnum(segind) +1; tmp = IRGB(r,c,:); tmp = tmp(:); histval(:,segind) = histval(:,segind) +tmp; end end histval = histval./repmat(histnum,3,1); qlv = 3; distv = sqdist(histval, histval); Y = distv; Y(conn==0)= inf; minY_col = zeros(1,n); minY_colsq = zeros(1,n); for i = 1:n j1 = 1:(i-1); j2 = (i+1):n; jtmp = [j1 j2]; ytmp = Y(i,jtmp); [tmp1,tmp2] = min(ytmp); minY_col(i) = tmp1; minY_colsq(i) = jtmp(tmp2); end R = 1:n;%n: number of vector Z = zeros(n-1,3); % allocate the output matrix. N = zeros(1,2*n-1); N(1:n) = histnum; centers = zeros(2*n-1,qlv); centers(1:n,:) = histval'; for s = 1:(n-1) % [v, k] = min(Y); ncur = length(minY_col); [v, k] = min(minY_col); ctmp = k; rtmp = minY_colsq(k); Z(s,:) = [R(rtmp) R(ctmp) v]; % update one more row to the output matrix A if s<n-1 % Update Y. i = min([rtmp ctmp]); j = max([rtmp ctmp]); I1 = 1:(i-1); I2 = (i+1):(j-1); I3 = (j+1):ncur; % these are temp variables U = [I1 I2 I3]; n1 = N(R(i)); n2 = N(R(j)); %updata conn t1 = conn(i,:); t2 = conn(j,:); t = t1|t2; conn(i,:) = t; conn(:,i) = t; tmp = conn(ncur,:); conn(j,:) = tmp; conn(:,j) = tmp; centers(n+s, :) = (n1*centers(R(i),:)+ n2*centers(R(j),:))/ (n1+n2); dtmp = (centers(R(U),:)- repmat(centers(n+s,:),length(U),1)); tmpY = sum(dtmp.^2,2);%merge min dist for two centriod? tmpY(conn(i,U)==0) = inf; Y(i,i) = 0; Y(i,U) = tmpY; Y(U,i) = tmpY; tmp = Y(ncur,1:ncur); Y(j,1:ncur) = tmp; Y(1:ncur,j) = tmp; [mintmpY,mintmpYsq] = min(tmpY); % minY_col(j)= minY_col(end); % minY_colsq(j)=minY_colsq(end); minY_col(i) = mintmpY; minY_colsq(i) = U(mintmpYsq); for sq = 1:length(tmpY) coltmp = U(sq); if (minY_colsq(coltmp)~=i)&&(minY_colsq(coltmp)~=j) if (tmpY(sq)<minY_col(coltmp)) minY_col(coltmp) = tmpY(sq); minY_colsq(coltmp) = i; end else %must update n1 = ncur-1; if coltmp==ncur % tmpY2 = Deall(1:n1,j)./DRall(1:n1,j); i0 = 1:(j-1); j0 = (j+1):n1; u0 = [i0 j0]; tmpY2 = Y(u0,j); else i0 = 1:(coltmp-1); j0 = (coltmp+1):n1; u0 = [i0 j0]; tmpY2 = Y(u0,coltmp); % tmpY2 = Deall(1:n1,coltmp)./DRall(1:n1,coltmp); end [mintmp1,mintmp1sq]=min(tmpY2); mintmp1sq = u0(mintmp1sq); minY_col(coltmp) = mintmp1; minY_colsq(coltmp) =mintmp1sq; end end minY_colsq(minY_colsq==ncur) = j; minY_col(j) = minY_col(end); minY_colsq(j) = minY_colsq(end); minY_col(end) = []; minY_colsq(end) = []; N(n+s) = N(R(i)) + N(R(j)); R(i) = n+s; R(j) = R(end); R(end) = []; end end T = cluster(Z,'maxclust',nseg );