matlab-(含教程)基于FMM+Criminisi算法彩色图像修复matlab仿真

function [thick_mask, thin_mask] = get_mask_2(I_gray, black_area) %检测图片二中栏杆的位置，返回两个mask，thick_mask是宽栏杆的mask，thin_mask是扁栏杆的mask BW=imbinarize(I_gray,0.15); %二值化 % figure; BW=((1-BW).*255); BW(black_area:end, :)=0; %去除下面的干扰部分 % imshow(BW);hold on; %检测宽栏杆 %找到宽栏杆内部的四个点 i_delta = black_area/4; i_points = floor(black_area/8):i_delta:floor(black_area*7/8); j_points = zeros(length(i_points)); count=1; neibor_len=floor(30*size(BW,1)/1210); for i=i_points j_neibor = zeros([1 size(BW,2)]); for j=1:size(BW,2) i_delta=min(min(neibor_len, i-1), size(BW,1)-i); j_delta=min(min(neibor_len, j-1), size(BW,2)-j); neibor_area = BW(floor(i-i_delta):floor(i+i_delta), floor(j-j_delta):floor(j+j_delta)); j_neibor(j)=sum(sum(neibor_area>0)); %i j领域中白点数量 end [~,j_point] = max(j_neibor); % plot(j_point, i_points(count),'x','LineWidth',2,'Color','red'); j_points(count)=j_point; count=count+1; end %从宽栏杆内部四个点分别向左向右出发，找到宽栏杆共八个边界点 count=1; for k=1:length(i_points) i_p = i_points(k); j_p = j_points(k); j_left=j_p;j_right=j_p; for j_search = j_p:-1:1 %找到宽栏杆的左边界点 if BW(floor(i_p), floor(j_search))==0 j_left=j_search; break; end end for j_search = j_p:1:size(BW,2) %找到宽栏杆的右边界点 if BW(floor(i_p), floor(j_search))==0 j_right=j_search; break; end end j_contour_left(count) = j_left; j_contour_right(count) = j_right; count = count+1; % plot(j_left, i_p,'x','LineWidth',2,'Color','yellow'); % plot(j_right, i_p,'x','LineWidth',2,'Color','yellow'); end %利用边界点拟合出宽栏杆的左右边界线 thick_line_left =polyfit(i_points, j_contour_left, 1); %左直线的k和b thick_line_right =polyfit(i_points, j_contour_right, 1);%右直线的k和b point_1_1 = 1*thick_line_left(1)+thick_line_left(2); point_1_2 = size(BW,1)*thick_line_left(1)+thick_line_left(2); point_2_1 = 1*thick_line_right(1)+thick_line_right(2); point_2_2 = size(BW,1)*thick_line_right(1)+thick_line_right(2); % plot([point_1_1 point_1_2], [1 size(BW,1)],'LineWidth',2,'Color','green'); % plot([point_2_1 point_2_2], [1 size(BW,1)],'LineWidth',2,'Color','green'); %得到宽栏杆左边区域和右边区域，并去除宽栏杆 oneRow = 1:size(I_gray,2); martix_row = repmat(oneRow, [size(I_gray,1) 1]); oneCol = 1:size(I_gray,1); martix_col = repmat(oneCol', [1 size(I_gray,2)]); ikLeft_b = martix_col.*thick_line_left(1)+thick_line_left(2); ikRight_b = martix_col.*thick_line_right(1)+thick_line_right(2); mask_left_left = martix_row<ikLeft_b; %在宽栏杆左直线左边的点为1 mask_right_right = martix_row>ikRight_b; %在宽栏杆右直线右边的点为1 mask_middle = and((martix_row>ikLeft_b),(martix_row<ikRight_b)); %在左右直线中间的点为1 mask_middle2 = and((martix_row>ikLeft_b-3),(martix_row<ikRight_b+3)); %拓展边界 BW(mask_middle)=0; thick_mask = mask_middle2*255; SE=strel('disk',2,4); BW_dil = imdilate(BW,SE); % figure;imshow(BW_dil); [lines] = get_hough_lines(BW_dil, 40); %直线检测找到直线 %将检测到的直线分成三部分：宽栏杆左边的横线，宽栏杆右边的横线，竖线 middle_line = (thick_line_left + thick_line_right)/2; left_lines = lines(1); %宽栏杆左边的横线 right_lines = lines(1); %宽栏杆右边的横线 up_lines = lines(1); %竖线 for line = lines point_mid=(line.point1+line.point2)/2; if abs(line.theta)<45 up_lines(length(up_lines)+1)=line; continue; end if line.point1(2)<5 && line.point2(2)<5 continue; end if point_mid(2)*middle_line(1)+middle_line(2)>point_mid(1) %左边直线 left_lines(length(left_lines)+1)=line; else right_lines(length(right_lines)+1)=line; end end if length(left_lines)>1 left_lines=left_lines(2:end); end if length(right_lines)>1 right_lines=right_lines(2:end); end if length(up_lines)>1 up_lines=up_lines(2:end); end %将检测到的直线标准化，转化成k、b形式，补充漏掉的直线，并去掉多余的直线 [left_new_lines] = lines_integrate(left_lines, 10, 15); left_new_lines = complete_lines(left_new_lines, 1, 1); [right_new_lines] = lines_integrate(right_lines, 10, 10); right_new_lines = complete_lines(right_new_lines, 11, 901*size(I_gray, 2)/1418); [up_new_lines] = lines_integrate(up_lines, 10, 10); % figure; % imshow(I_gray); hold on; % for i=1:length(up_new_lines) % line=up_new_lines(i); % p1=[0 -line.b/line.k]; % p2=[size(I_gray,2) (size(I_gray,2)-line.b)/line.k]; % plot([p1(1) p2(1)], [p1(2) p2(2)],'LineWidth',2,'Color','green'); % end %利用直线位置得到mask line_width=floor(4*size(I_gray,1)/1210); [left_line_mask] = maskFromlines(I_gray, line_width, left_new_lines, black_area, 6, false); left_line_mask = left_line_mask.*mask_left_left; [right_line_mask] = maskFromlines(I_gray, line_width, right_new_lines, black_area, 13, false); right_line_mask = right_line_mask.*mask_right_right; [up_line_mask1] = maskFromlines(I_gray, line_width*1.7, up_new_lines(1), black_area, 8, true); [up_line_mask2] = maskFromlines(I_gray, line_width, up_new_lines(2), black_area, 6, true); up_line_mask=up_line_mask1+up_line_mask2; %返回两个mask，thick_mask是宽栏杆的mask，thin_mask是扁栏杆的mask thin_mask = right_line_mask+left_line_mask+up_line_mask; thin_mask(thin_mask>1)=1; thin_mask=thin_mask.*255; thin_mask(floor(391*size(I_gray,1)/605):end, :) = 0; I_gray_temp = I_gray; I_gray_temp(thick_mask>0)=255; % figure;imshow(I_gray_temp); end %根据直线间的距离补充漏掉的直线 function [lines]=complete_lines(lines, add_num, left_start_y) delta_m = lines(2).m-lines(1).m; i=2; while true if i==length(lines) break; end if (lines(i+1).m-lines(i).m)>1.5*delta_m %中间缺一根线 middle_m = lines(i).m+delta_m; middle_k = lines(i).k; middle_b = 1-middle_k*middle_m; lines(i+2:end+1)=lines(i+1:end); lines(i+1).m = middle_m; lines(i+1).k = middle_k; lines(i+1).b = middle_b; else delta_m = lines(i+1).m-lines(i).m; end i=i+1; end start_m = (left_start_y-lines(end).b)/lines(end).k; for i=1:add_num start_m = start_m+delta_m; %最后新增一条横线 middle_k = lines(end).k; middle_b = left_start_y-middle_k*start_m; count = length(lines)+1; lines(count).m = (1-middle_b)/middle_k; lines(count).k = middle_k; lines(count).b = middle_b; end end