期刊名称：International Journal of Software Engineering and Its Applications
印刷版ISSN：1738-9984
出版年度：2014
卷号：8
期号：2
页码：409-416
DOI：10.14257/ijseia.2014.8.2.39
出版社：SERSC
摘要：In this paper, we propose a novel method for automatic object segmentation using Mean shift with Grid-mask for Grab Cut algorithm. The main idea of proposed method is a rapid and exact technique for extracting initial foreground information using Mean shift with Grid-mask then we make initial rectangle around estimated initial foreground. And then Grab Cut algorithm is applied to segment foreground form background based on initial rectangle provided by previous process. In order to evaluate our proposed method, we compare the proposed method with several competitive automatic methods on the MSRA database which has 1000 images with ground truth. The scheme successfully segments objects without prior knowledge with both higher precision and better recall than competitive methods.
