Region Segmentation by Virtual Super Pixel and Global Features

Abstract

A region segmentation method developed by combining the virtual edge extraction method and the region segmentation method based on global features is presented in this paper. In the concrete, small blocks called "virtual super pixel" which are obtained by the virtual edge extraction method are used are used as an atomic element of the region integration in the present paper. Not only the deficit of the conventional region segmentation based on global features such that the thin line like region is not extracted is conquered, but also the process is simplified.