Remove White Pixels from Image for Processing
显示 更早的评论
I have some tissues stained with Masson's Trichrome. I have posted a sample picture below.
I am attempting to quantify the amount of blue stain as a fraction of the amount of red stain + blue stain in this image, and images like it. To accomplish this, I use a k-means algorithm (im is just the image in this case).
My only issue with this algorithm is that it outputs the percent of the total image that is blue, when I want it to just output the percent of the tissue that is blue. I would like to threshold out, based on brightness, the background pixels. Ideally, this would be before the k-means clustering algorithm would take place, so as to avoid any interference that background has on the algorithm. Unfortunately, I am a novice in matlab, and have no idea how I would do any of this. If any of you can provide some help, I'd greatly appreciate it.
im = imread(fullfile(obj.pathlist{i, 1}, obj.pathlist{i, 2}));
makecform('srgb2lab');
[seg_im, PercentStained(i)] = QuantifyArea.blueThreshold(im, cform);
function [seg_im, percent_stained] = blueThreshold(im, cform)
%BLUETHRESHOLD performs thresholding for Masson's Trichrome
%stain
%convert colours
lab_he = applycform(im,cform);
%isolate the colour and luminance data.
ab = double(lab_he(:, :, 2:3));
L = double(lab_he(:, :, 1));
%use a k-means clustering algorithm to isolate'red' and 'blue'
nrows = size(ab,1); ncols = size(ab,2);
ab = reshape(ab,nrows*ncols,2);
[cluster_idx, cluster_center] = kmeans(ab, 2,'distance','sqEuclidean', 'Replicates',3);
%order the cluster labels based on the cluster centers.
[~, sorted_cluster_index] = sort(cluster_center(:, 1), 1, 'ascend');
%isolate the red and blue regions
pixel_labels = reshape(cluster_idx,nrows,ncols);
blue_image = (pixel_labels == sorted_cluster_index(1));
red_image = (pixel_labels == sorted_cluster_index(2));
%don't include blue voxels that are within red regions, as they are
%likely noise
strel1 = strel('disk', 3, 8);
red_image = imdilate(red_image, strel1);
blue_image = imclose(blue_image, strel1);
%create an image overlay for quality assurance and calculate the percent
%blue voxels
blue_image = uint8(repmat(blue_image.*(red_image == 0 & L <= 210 & L >= 0), [1 1 3]));
seg_im = blue_image.*uint8(permute(repmat([0; 0; 255], [1, nrows, ncols]), [2 3 1])) + uint8(blue_image == 0).*uint8(permute(repmat([255; 0; 0], [1, nrows, ncols]), [2 3 1]));
%calculate the percent area of the image stained blue
percent_stained = sum(blue_image(:))/(3*nrows*ncols)*100;
end
回答（1 个）
darova
2019-3-5
Doesnt work fast, but you can separate any color you want:
clc, clear
A = imread('images\masson.jpg');
subplot(1,2,1)
imshow(A)
[m, n, ~] = size(A);
RGB = [190 226 250]; % color you want
e = 20; % color shift
B = zeros(m,n,3);
B = uint8(B)+255;
count = 0;
for i = 1:m
for j = 1:n
if pix_in(A(i,j,:), RGB, e)
count = count + 1;
B(i,j,:) = A(i,j,:);
end
end
end
subplot(1,2,2)
imshow(B)
percent = count/(m*n)*100;
% And the function
function b = pix_in(A,RGB,e)
B = [0 0 0];
for k = 1:3
t = double(A(1,1,k));
B(k) = abs(t-RGB(k)) < e;
end
if sum(B) == 3, b = true;
else b = false;
end
end
类别
在 帮助中心 和 File Exchange 中查找有关 Images 的更多信息
2019-3-4
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
Translated by 
