[Project] Cell segmentation and counting based on matlab watershed algorithm [with Matlab source code 639]

1. Introduction

The watershed algorithm is an image region segmentation method. During the segmentation process, the picture is converted into a grayscale image. Then I will regard the grayscale value as the altitude, and then inject water into the lower point. This interpretation based on topography, we Three points are emphatically considered: the

minimum point, which corresponds to the lowest point of a basin. When we drop a drop of water in the basin, the water will eventually converge to this point due to gravity. Note: There may be a minimum surface, and all points in this plane are minimum points.
At other locations in the basin, the droplets from that location will converge to the local minimum.
The edge point of the basin is the junction of the basin and other basins. A drop of water at this point will flow to any basin with equal probability.

After understanding the above three points, we start to inject water into the minimum points of the basin, and then as the water injection deepens, each minimum point slowly expands outwards, and then we know that the water in the two basins converges, and the confluence is The watershed we need.

It can be understood intuitively from the figure below. First, these three areas contain minimum points,

and then fill them gradually to obtain the watershed (ie the dividing line) to

obtain the dividing line to complete the image segmentation:

Second, the source code

function susanseg
clear all; close all; clc
image= imread('cell.jpg');
% 用SUSAN算法进行边缘检测
image = susan(image,4);
figure, imshow(image,[]);
%imwrite(image, './susanout/susanout.jpg');
% 将image转为二值图像保存后，用图像处理工具
% 把其背景的所有连通区域处理为黑色，即只有细
% 胞体是白色，便于细胞数目的搜索
BW = im2bw(image, graythresh(image));
bounder_area = length(find(BW==0));
%imwrite(BW, './susanout/bw.jpg');
figure, imshow(BW);


% 申明全局变量
global B Dir m n;
B = imread('./blackbackground.jpg');
B = im2bw(B, graythresh(B));
[m,n] = size(B);
figure, imshow(B);

% 细胞的总面积，即细胞所占的像素数目，包括细胞的边界
% 由于SUSAN提取出的边界已被增宽，所以将边界像素数除以2
% 来作为细胞的边界像素数目
total_area = length(find(B==1)) + bounder_area/2;
NUM = 5; % 细胞面积阈值
count = 0; % 细胞总数
% 搜索方向向量，4邻域搜索
Dir = [-1 0; 0 1; 1 0; 0 -1;];
% 搜索方向向量，8邻域搜索
%Dir = [-1 0; -1 1; 0 1; 1 1; 1 0; 1 -1; 0 -1; -1 -1;];
for i = 1:m
    for j = 1:n
        if B(i,j)==1 % 是细胞像素
            num = search(i,j,4) + 1; % 计算该细胞的像素数目
            if num>NUM
                count = count  + 1;
            else
                total_area = total_area - num; % 减掉不是细胞的面积
            end
        end
    end
end
%fid = fopen('./susanout/results.txt', 'wt');
fprintf('图像尺寸: %d * %d, SUSAN阈值: 4, 细胞面积阈值: %d\n', ...
    n, m, NUM);
fprintf('细胞总数: %d, 细胞总面积: %.2f, 平均细胞面积: %.2f\n', ...
    count, total_area, total_area/count);
%fprintf(fid,'图像尺寸: %d * %d, SUSAN阈值: 4, 细胞面积阈值: %d\n', ...
%    n, m, NUM);
%fprintf(fid,'细胞总数: %d, 细胞总面积: %.2f, 平均细胞面积: %.2f\n', ...
%    count, total_area, total_area/count);
%fclose(fid);
end
% -----------------------------------------------------------------------
% 
% This function uses the SUSAN algorithm to find edges within an image
% 
%
% >>image_out = susan(image_in,threshold)
%
%
% Input parameters ... The gray scale image, and the threshold 
% image_out .. (class: double) image indicating found edges
% typical threshold values may be from 10 to 30
%
%
%The following steps are performed at each image pixel: 
% ( from the SUSAN webpage, http://www.fmrib.ox.ac.uk/~steve/susan/susan/node4.html )
% 
% Place a circular mask around the pixel in question. 
% Calculate the number of pixels within the circular mask which have similar brightness to 
% the nucleus. These define the USAN. 
% Subtract USAN size from geometric threshold to produce edge strength image. 
%
% Estimating moments to find the edge direction has not been implemented . 
% Non-maximal suppresion to remove weak edges has not been implemented yet.
%
% example:
%
% >> image_in=imread('test_pattern.tif');
% >> image = susan(image_in,27);
% >> imshow(image,[]) 
%
%
% Abhishek Ivaturi
% 
% -------------------------------------------------------------------------
 

function image_out = susan(im,threshold)

% check to see if the image is a color image...
%im= imread('test_pattern.tif')
%threshold=27;
d = length(size(im));
if d==3
    image=double(rgb2gray(im));
elseif d==2
    image=double(im);
end

% mask for selecting the pixels within the circular region (37 pixels, as
% used in the SUSAN algorithm

mask = ([ 0 0 1 1 1 0 0 ;0 1 1 1 1 1 0;1 1 1 1 1 1 1;1 1 1 1 1 1 1;1 1 1 1 1 1 1;0 1 1 1 1 1 0;0 0 1 1 1 0 0]);  


% the output image indicating found edges
R=zeros(size(image));


% define the USAN area
nmax = 3*37/4;

% padding the image
[a b]=size(image);
new=zeros(a+7,b+7);
[c d]=size(new);
new(4:c-4,4:d-4)=image;
  
for i=4:c-4
    
    for j=4:d-4
        
        current_image = new(i-3:i+3,j-3:j+3);
        current_masked_image = mask.*current_image;
   

%   Uncomment here to implement binary thresholding

%         current_masked_image(find(abs(current_masked_image-current_masked_image(4,4))>threshold))=0;         
%         current_masked_image(find(abs(current_masked_image-current_masked_image(4,4))<=threshold))=1;


%   This thresholding is more stable
                 
                   
        current_thresholded = susan_threshold(current_masked_image,threshold);
        g=sum(current_thresholded(:));
        
        if nmax<g
            R(i,j) = g-nmax;
        else
            R(i,j) = 0;
        end
    end
end

Three, running results

Four, remarks

Complete code or write on behalf of adding QQ 1564658423