Segmentace obrazu metodou "mean shift" a její efektivní implementace

Abstract

An image segmentation is one of the fundamental tasks of computer vision. One of the commonly used segmentation techniques is a method known as ”meanshift”. It is a robust method that requires only small number of user parameters. One of the important parameters is a bandwidth matrix. The aim of this thesis was to implement the filtering part of meanshift method and study its behavior with different settings of the bandwidth matrix. This behavior has been tested on real images, and especially the white Gaussian noise.