min max flow graph cuts for image segmentation MATLAB wrapper

802 views Asked by At

My objective is to segment to perform cell segmentation. The cells are of different sizes, and the image is grayscale using graph cuts I used the MATLAB wrapper by Shai Bagon, and I am able to run the basic test code which was given here

I am not able to get the label image as desired. I need help to interpret the results from this method.

Also another quick question, from the examples I have looked do far. Graph cut seems to be used for images with a single connected foreground and the rest is background. However I can not find examples where the foreground is disjoint (as in my case, cells all over the image) and the background. This makes me assume that graph cuts does not work on a disjoint foreground ? If someone can let me know this is correct.

Your help is appreciated. Thank you

1

There are 1 answers

1
Shai On BEST ANSWER

Regarding your results on cell segmentation: you provide very little input to work with. It is unclear what you are trying to do and what is wrong with what you actually get. You'll have to post some example input image and short code.

As for the second question, in general GraphCut image segmentation is not restricted to a single connected output segment. The number of segments depend on the data and smoothness terms provided. As a rule of thumb, the weaker the smoothness term the more connected-components you'll see at the output.