How to keep only solid white dots and eliminate dots that inside them is black in the attached image?

Hello, Is there any way that I could remove the dots which are black inside from the attached image to just keep the solid white dots in the image? Thanks

Answer 1

[Akira Agata]

bwpropfilt function would be helpful for your task. The following is an example.

% Read the image
I = imread('Example.jpg');

% Binarize the image
BW = imbinarize(rgb2gray(I));

% Suppress light structures connected to image border
BW = imclearborder(BW);

% Filter out regions which has hole(s)
BW = bwpropfilt(BW,'EulerNumber',[1 1]);

[Penny13]

Thank you so much. It worked and deleted the holes. So I got the image Example2.jpg. Now could you also guide me on how I could eliminate every none solid shape too? If you look at the Example2, there are some irregular shapes in there.

[Akira Agata]

Could you upload Example2.jpg again? Anyway, how to eliminate none solid shape depends on how you define "non solid shape". For example, if you want to eliminate C-like shape, checking on ratio of 'Area' to 'Perimeter' (or 'Area' to 'ConvexArea') might be one possible solution.

[Penny13]

Thanks a lot for your answer. If you don't mind I will ask one last question. For every dot that I have in this image Example2.jpg, How could I draw a square with only having the centroid of the dot?

[Akira Agata]

Please use the regionprops function to detect centroid and bounding box for each area. The following is an example.

% Read the image
I = imread('Example2.jpg');

% Binarize the image
BW = imbinarize(rgb2gray(I));

% Suppress light structures connected to image border
BW = imclearborder(BW);

% Filter out regions which has hole(s)
s = regionprops(BW,{'Centroid','BoundingBox'});

% Show the result
figure
imshow(BW)
hold on
for kk = 1:numel(s)
  scatter(s(kk).Centroid(1),s(kk).Centroid(2),'rx')
  rectangle('Position',s(kk).BoundingBox,'EdgeColor','y')
end

Since white areas are relatively small against image size, the result becomes like this.

But if you expand part of the image, you can see the image like this: