A new semi-supervised method for image co-segmentation

Image co-segmentation addresses the problem of simultaneously extracting the common targets from a set of related images. However, designing a robust and efficient co-segmentation algorithm is a challenging work because of the variety and complexity of the object and the background. In this paper, we propose a new semi-supervised method to extract foreground object from an image collection. The proposed method is composed of three tasks: 1) object proposal generation, 2) object prior propagation and 3) foreground extraction. The main idea of this paper is to transfer the segmentation from a subset of training images to test images. The comparison experiments conducted on public datasets iCoseg and MSRC demonstrate the performance of the proposed method.