Research of color image segmentation algorithm based on asymmetric kernel density estimation

Abstract

Traditional kernel density estimation method only depends on a given sample, in which the same weight of kernel function reduces the ability to distinguish the category of image region, though it has certain advantages in image segmentation. A novel method based on asymmetric kernel density estimation is introduced for more accurately integrating the differences among color features of samples marked by users. The method differently treats the kernel function, and the weight coefficient is introduced in the kernel density estimation function to express each kernel function's contribution to the overall estimation. Simulation experimental results show that our proposed method is more powerful in category description and distinguishing, which enhances the regional information constraints and robustness of the segmentation model and the integrity of the target region, and more accurately segments thin elongated region when compared with traditional kernel density estimation method.