Image Segmentation using Hierachical Feature Selection (HFS)

The algorithm is executed in 3 stages. In the first stage, it obtains an over-segmented image using SLIC. In the last 2 stages, it iteratively merges the over-segmented image with merging method mentioned in EGB (Efficient Graph-based Image Segmentation) and learned SVM parameters.

Sources:

https://github.com/opencv/opencv_contrib/blob/3.4.1/modules/hfs/samples/example.cpp

read an image

img = imread(fullfile(mexopencv.root(),'test','fruits.jpg'));

initialize object with image height and width

obj = cv.HfsSegment(size(img,1), size(img,2));
obj.SlicSpixelSize = 8;  % change any parameter you want

perform segmentation

out = obj.performSegment(img, 'Draw',true);
imshow(out)

instead of getting a segmented image, we can also get a matrix of indices

L = obj.performSegment(img, 'Draw',false);
figure, imshow(label2rgb(L))