A Comparative Study of KMCG Segmentation Based on YCbCr, RGB, and HSV Color Spaces

Abstract

Kekre Median Codebook Generation (KMCG) is a vector quantization algorithm. It is used for several purposes like image compression and segmentation. It has been applied by several application and shows its efficiency. This paper presents a comparison study of applying KMCG with three color models: RGB, YCbCr, and HSV for image segmentation. The experiments applied on five images, three of them are benchmarks. Two numerical metrics are utilized: E measure and Peak Signal to Noise Ratio (PSNR), in addition to the visual results. The final results show that KMCG conducts better segments when it is applied with the RGB color model. It returns more homogenies segments than using KMCG with YCbCr or HSV.