mean_shift算法MATLAB代码

function [k_hist,target_hist,khist] = compute_k_hist(frame,target,kmatrix,MM) % 函数和文件名 ： color_compute_k_hist % 版本 ： 1.0 % 创建时间 ： 2007.5.30 % 作者 ： pineapple % 实现功能 ： 计算彩色图像核直方图，和灰度图一样，只是级数=mm*mm*mm % 输入： % frame : 当前帧 % target ： 选定的目标区域 % kmatrix ： 核矩阵 % MM : 量化级数 % 输出： % target_hist ： 目标target的mm级直方图矩阵 % k_hist ： 目标target的核直方图 % khist ： 分块后（四小块的各自的核直方图） % % 注意： % 为了和分块检测功能相符合，这里把目标分成四块分别计算其分布。 % khist是一个结构体 % ********************************************************************** % [m , n , k] = size(frame); if(k==3) if nargin < 4 mmr = 8; mmg = 8; mmb = 8; end yyyy=mmr*mmg*mmb; khist.up = zeros(1,yyyy); %上 khist.right = zeros(1,yyyy); % khist.left = zeros(1,yyyy); % khist.down = zeros(1,yyyy); % parts_k_hist = zeros(9,mmr*mmg*mmb); cpoint = target.cpoint; width = target.width; k_hist = zeros(1,yyyy); target_image = frame(cpoint(1)-width(1):cpoint(1)+width(1),cpoint(2)-width(2):cpoint(2)+width(2),:); [m,n,k] = size(target_image); target_hist = zeros(m,n); target_image = double(target_image); parts_index = [1 (m-1)/2 1 (n-1)/2; %左上 1 1 (m-1)/2 (n+3)/2 n; %右上 2 (m+3)/2 m (n+3)/2 n; %右下 3 (m+3)/2 m 1 (n-1)/2; %左下 4 (m+1)/2 (m+1)/2 1 (n-1)/2; % 中间行，左列 5 (m+1)/2 (m+1)/2 (n+3)/2 n; % 中间行，右列 6 1 (m-1)/2 (n+1)/2 (n+1)/2 ; % 中间列，上行 7 (m+3)/2 m (n+1)/2 (n+1)/2 ; % 中间列，下行 8 (m+1)/2 (m+1)/2 (n+1)/2 (n+1)/2] ; % 中间列，下行 9 for k = 1 : 9 for i = parts_index(k,1) :parts_index(k,2) for j = parts_index(k,3) :parts_index(k,4) temp_1 = 1 + fix(target_image(i,j,1)*mmr/256); temp_2 = 1 + fix(target_image(i,j,2)*mmg/256); temp_3 = 1 + fix(target_image(i,j,3)*mmb/256); value = (temp_1-1)*mmg*mmb+(temp_2-1)*mmb+temp_3; target_hist(i,j) = value; parts_k_hist(k,value) = parts_k_hist(k,value) + kmatrix(i,j); k_hist(1,value) = k_hist(1,value) + kmatrix(i,j); end end end % 归一化 k_hist = k_hist./(sum(k_hist)); khist.rigth = parts_k_hist(2,:) + parts_k_hist(3,:) + parts_k_hist(7,:)+ parts_k_hist(8,:)+ parts_k_hist(9,:)+ parts_k_hist(6,:); khist.left = parts_k_hist(1,:) + parts_k_hist(4,:) + parts_k_hist(7,:)+ parts_k_hist(8,:)+ parts_k_hist(9,:)+ parts_k_hist(5,:); khist.up = parts_k_hist(1,:) + parts_k_hist(2,:) + parts_k_hist(5,:)+ parts_k_hist(6,:)+ parts_k_hist(9,:)+ parts_k_hist(7,:); khist.down = parts_k_hist(3,:) + parts_k_hist(4,:) + parts_k_hist(5,:)+ parts_k_hist(6,:)+ parts_k_hist(9,:)+ parts_k_hist(8,:); khist.rigth = khist.rigth./(sum(khist.rigth)); khist.left = khist.left./(sum(khist.left)); khist.up = khist.up./(sum(khist.up)); khist.down = khist.down./(sum(khist.down)); end % cpoint = target.cpoint; % width = target.width; % k_hist1 = zeros(1,mmr*mmg*mmb); % target = frame(cpoint(1)-width(1):cpoint(1)+width(1),cpoint(2)-width(2):cpoint(2)+width(2),:); % [m,n,k] = size(target); % target_hist1 = zeros(m,n); % target_image = double(target); % for i =1:m, % for j=1:n % temp_1 = 1 + fix(target_image(i,j,1)*mmr/256); % temp_2 = 1 + fix(target_image(i,j,2)*mmg/256); % temp_3 = 1 + fix(target_image(i,j,3)*mmb/256); % value = (temp_1-1)*mmg*mmb+(temp_2-1)*mmb+temp_3; % target_hist1(i,j) = value; % k_hist1(1,value) = k_hist1(1,value) + kmatrix(i,j); % end % end % sum(k_hist1) % sum(kmatrix(:)) % k_hist - k_hist1 % target_hist1 - target_hist