Improved Grab Cut with Human Visual Perception

Grab Cut can effectively extract the foreground from cartoon images; however, it is poor for real images. To improve the segmentation performance, we propose an improved Grab Cut based on human visual perception, which is formulated as a joint optimization for segmentation and visual appearance parameters. In this model, we represent the image as a pseudo-color image in order to classify the pixels with identical visual perception into a region, and use the histogram shape analysis method to estimate the optimal number of regions for the foreground and background. Experimental results indicate that the proposed method is competitive in terms of segmentation metrics and computational expense.