How to remove unconnected regions from image

Hello

I am doing project "Persons identification on the basis of retinal vessels". I am working on the part of segmentation of blood vessels. I have read some relevant papers to implement segmentation process. However segmentation is not giving good results with majority of my images. I have to get good segmentation result for all the images. Some of papers I read are: <http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=6965041> I also used information on the link https://www.mathworks.com/matlabcentral/answers/113073-how-to-remove-unconnected-pixels-or-objects-from-an-image and solved some problems using bwareafilt() and bwareaopen(). Below is my one of the input image with which segmentation is not good at all. Please help me with this. I also attached my .m file. Thanks in advance.

[Image Analyst]

That is an extraordinarily bad image. I've never seen a fundus image with such low contrast and strange colors. Did something go wrong? Or is there something truly anatomically wrong with that subject? I'd ask your ophthalmologist if there was some problem with the image capture settings.

[Sidra Aleem]

Actualy this image is in my data set. With the normal image I am getting very good result. So can you sugest me some solution for such kind of images.As there are many such images in my data set. And have ti make my program robust.

[Image Analyst]

Sorry, developing an algorithm to handle such bad images would take more time that I can devote to a question. It could take months or years. You might check the literature http://www.visionbib.com/bibliography/contentsmedical.html#Medical%20Applications,%20CAT,%20MRI,%20Ultrasound,%20Heart%20Models,%20Brain%20Models or hire a university or think tank to help write the algorithm for you.

[Sidra Aleem]

I contacted doctor who provided me data set. She said that some of the images are effected by catract and mayopia. So its natural anatomical structure.

[Image Analyst]

What is she hoping that you can visualize or measure that she can't see without you? Is she willing to manually trace or outline things?

Answer 1

[sam CP]

You can refer the following code..

binaryImage = bwareafilt(binaryImage, 2); % Extract 2 largest blobs. % Erode it a little with imopen(). figure, imshow(binaryImage,[]);title('bwareafilt2'); binaryImage = imopen(binaryImage, true(5)); figure, imshow(binaryImage,[]);title('imopen');

binaryImage = bwareafilt(binaryImage, 1); % Extract largest blob. figure, imshow(binaryImage,[]);title('bwareafilt1'); % Fill any holes in the lagest blob. binaryImage = imfill(binaryImage, 'holes'); figure, imshow(binaryImage,[]);title('imfill'); % Dilate mask out a bit in case we've chopped out a little bit of largest blob. binaryImage = imdilate(binaryImage, true(5)); figure, imshow(binaryImage,[]);title('imdilate');

% Display the final binary image.

figure, imshow(binaryImage, []);title('Final binary image');

%===========================================================================================================

[Sidra Aleem]

No that dint work for me.