A Review on Image Co-segmentation

In this paper, the current image co-segmentation methods are reviewed and summarized. According to the characteristics of co-segmentation problem, the formal definition to the problem is given. Consequently, the co-segmentation methods are divided into two categories: single-object based co-segmentation and multi-objects based. The main methods of these two categories are introduced and analyzed. From the four aspects of segmentation model construction, feature selection, co-information usage and optimization method, we explore the similarities and differences between the methods, as well as their advantages. We further summarize the challenges and development directions of image co-segmentation for benefiting the follow-up studies.