An image segmentation algorithm using iteratively the mean shift

Abstract

Image segmentation plays an important role in many systems of computer vision. The good performance of recognition algorithms depend on the quality of segmented image. According to the opinion of many authors the segmentation concludes when it satisfies the observer's objectives, the more effective methods being the iterative, However, a problem of these algorithms is the stopping criterion. In this work the entropy is used as stopping criterion in the segmentation process by using recursively the mean shift filtering. In such sense a new algorithm is introduced. The good performance of this algorithm is illustrated with extensive experimental results. The obtained results demonstrated that this algorithm is a straightforward extension of the filtering process. In this paper a comparison was carried out between the obtained results with our algorithm and with the EDISON System [16]. © Springer-Verlag Berlin Heidelberg 2006.