A fast and effective algorithm for image co-segmentation

Abstract

The target of co-segmentation is to recognize the same object among many images and segment the object [1]. Generally, the model of co-segmentation is a Markov random field (MRF) model [1][2]. The complexity of optimizing the problem is NP-hard [2]. Actually, there is a high-speed approximate algorithm that can solve the problem well, this paper proposed an approximate algorithm which combines the color and shape information and image segmentation preprocessing to simplify the problem. Experiment results show that the algorithm has a high successful rate and its low error rate is also acceptable under the circumstance of its high speed. © 2012 Springer-Verlag GmbH.