How to segment an image of an object from it body with the presence of shadow with same or lower intensity value?

clear all; close all; clc;
img = imread("enhancedimage.jpg");
I1 = rgb2gray(img);
figure, imshow(I1);
% imageSegmenter(I1);

% function [BW,maskedImage] = segmentImage(I1)
% %segmentImage Segment image using auto-generated code from imageSegmenter app
% %  [BW,MASKEDIMAGE] = segmentImage(X) segments image X using auto-generated
% %  code from the imageSegmenter app. The final segmentation is returned in
% %  BW, and a masked image is returned in MASKEDIMAGE.
% 
% % Auto-generated by imageSegmenter app on 06-Nov-2023
% ----------------------------------------------------


% Adjust data to span data range.
X1 = imadjust(I1);

% Threshold image - manual threshold
BW = X1 > 10;

% Active contour
iterations = 10;
BW = activecontour(X1, BW, iterations, 'Chan-Vese');

% Create masked image.
maskedImage = X1;
maskedImage(~BW) = 0;
% end
figure, imshow(maskedImage);
C=~maskedImage;
figure, imshow(C);

As you can see from the figure attached, I am trying to segment a nucleus (rounded-like shape) from a cell with a presence of shadow on the cell. I have try thresholding, Otsu thresholding and circular hough Transform. The result gave the same which are including the shadow and dark background as an object. Is there a way to segment the nucleus only? I have try to control the threshold value but it did not meet the expectation.

A comparison of the grayscale image and it segmentation

Answer 1

[Image Analyst]

First you can ask bwareafilt to give you only blobs that fall into a certain size range (range of area in pixels).

mask = bwareafilt(mask, [minArea, maxArea]); % Get just acceptable areas.

Then you can ask regionprops to give you the circularities. From those you can distinguish roundish objects from oblong or stick-like objects. Something like

labeledImage = bwlabel(mask); % Give each blob an ID number.
props = regionprops(labeledImage, 'Circularity');
allCircs = [props.Circularity]
goodCircs = allCircs > 0.6; % Or whatever.  This is a logical vector of true or false.
goodIndexes = find(goodCircs); % Convert logical vector to actual index numbers.
% Extract only the good blobs.
mask = ismember(labeledImage, goodIndexes);

For a full tutorial, see my Image Segmentation Tutorial in my File Exchange:

Image Analyst's File Exchange

It's a generic, general purpose demo of how to threshold an image to find blobs, and then measure things about the blobs, and extract certain blobs based on their areas or diameters.

[Khalis]

Thank you!