Image segmentation based on g-un-mms and heuristics. Theoretical background and results

Abstract

We present the concept of image segmentation based on Gauss - Uniform Noise Mixed Models (GUN-MM) that we also introduce in this paper, the theoretical foundation of the segmentation method based on this model, results of the method on several classes of pictures, and a comparison with other methods. The proposed implementation of the G-UN-MM method has a simple and sound theoretical foundation and is not computationally demanding. It produces in many cases better segmentation results than other methods that are more computationally intensive.