Study Of Segmentation In Color Remote Sensing Images

This paper deals with segmentation of color remote sensing images (CRSI). Researches on segmentation are performed upon CRSI from urban areas as well as from rural areas, respectively. A segmentation method of CRSI from urban areas based on mathematical morphology is proposed here. The segmentation is achieved through an improved color fuzzy probabilistic model, which is built upon the characteristic size and contrast of each structure (building and vegetation) measured by mathematical morphology. The other proposed segmentation method in this paper is about CRSI from rural areas which is characterized by strong texturally. The segmentation is achieved by using the textural features which are provided by the definitions of texture direction and matrix norm, as well as the unitary color. The experiments results show that the two methods proposed in this paper enjoy good performances on segmentation of CRSI.