Object cosegmentation by nonrigid mapping

• We use nonrigid knowledge mapping method to estimate the foreground mask.
• By using the deformable part models in the detecting process, we modify the detection process in traditional multiple foreground cosegmentation algorithm.
• We use several useful features in representing the object-based detecting window.

Image segmentation is an important research topic in image processing and computer vision. Recently, cosegmentation has received more and more attention. Although lots of research efforts have already studied this problem in the case of single object, there still lacks the deep investigation on multiple objects cosegmentation. In this paper, we try to attack this challenge by transferring the foreground segmentations using nonrigid mapping. We present a framework, in which we first take advantage of deformable part models to detect the foreground regions across the images, and the segmentation is formulated as an energy minimization problem on pixel labeling. We have conducted a set of experiments on the FlickrMFC dataset and iCoseg dataset. The experimental results demonstrate that our proposed approach outperforms the state-of-the-art methods.