Reputation: 1624
I begin with the following image:
I perform local thresholding by comparing a pixel to its neighborhood mean. I also crop out the white border surrounding the image after this local thresholding with the following code:
I = imread('img_path');
N = 21;
localMean = conv2(double(I), double(1/(N^2) * ones(N)), 'same');
BW = I > localMean;
borderSize = ceil(double(N) / 2);
R = [borderSize, borderSize, size(BW, 2) - 2*borderSize, size(BW, 1) - 2*borderSize];
BW = imcrop(BW, R);
This leaves me with the following image:
As you can see, there is a clear black border surrounding each of the segmented cells. How do I go about removing all things in this image OTHER than those regions within the black border?
If this is not possible, then are there any other techniques I should take a look at that might make give my local thresholding less noise?
Any help is appreciated.
Upvotes: 0
Views: 74
Reputation: 4195
you can use Otsu's thresholding (graythresh
) and morphological operations:
im = im2double(imread('cells.png'));
% otsu thresholding
t = graythresh(im);
bw = im > t;
% morphological open to remove small noise
bw = imopen(bw,strel('disk',5,0));
% fill holes to form solid objects
bw = imfill(bw,'holes');
% show
imshow(bw)
Upvotes: 1