An improved fast mean shift algorithm for segmentation

摘要

The mean shift algorithm is a statistical iterative algorithm based on kernel density estimation which has been widely used in many fields. This paper improves the mean shift algorithm by adopting the following approaches. Firstly, we present a novel approach named Random Sampling with Contexts (RSC) to speed up the mean shift algorithm. Secondly, we introduce Dempster-Shafer (D-S) theory for the fusion of features to improve the segmenting quality. Moreover, experimental results show that the new algorithm is superior to the typical mean shift algorithm.