Improving Comprehensive Sampling Sets Matting Using Texture Feature
Xin Hong, Yingyun Yang, Shuhong Wen · 2018
Digital matting refers to the precise separation of the foreground from the background of image or video. It is a key technology in the fields of image processing and video editing which has a very broad application prospect. In this paper, we improve the two most important problems to be solved in the sampling based matting algorithms-how to handle unknown pixels that fail to acquire foreground and background pixel samples and how to process the color distribution overlapped regions of the foreground and the background pixels. By adding the LBP texture cost function to the CSS algorithm, a more comprehensive and representative samples are extracted, and then we can select the best foreground and background samples from them. The experimental results show that the algorithm in this paper is more comprehensive in sample collection, and the result of matting is more accurate.