Saliency Detection Via Markov Chain

In this paper, we firstly review the principle of Markov chain, the concept of partially absorbing random walk, and briefly introduce the robust background prior. Then we respectively formulate saliency object detection via the absorbed time in absorbing Markov chain, the hitting time in ergodic Markov chain, the absorbing probability in partically absorbing random walk. Finally, we combine robust background prior with the obtained saliency maps into a unified framework to optimize the maps.We jointly consider the appearance divergence and spatial distribution of salient objects and the background. The virtual boundary nodes are chosen as the absorbing nodes in a Markov chain and the absorbed time from each transient node to boundary absorbing nodes is computed. The absorbed time of transient node measures its global similarity with all absorbing nodes, and thus salient objects can be consistently separated from the background when the absorbed time is used as a metric. We discard the time staying in starting node in absorbed time to automatically eliminate the outliers.We analysis and compare the results of saliency detection via the absorbed time in absorbing Markov chain and the hitting time in ergodic Markov chain. We conclude that the absorbed time can synthetically consider the relationships between a node and multiple absorbing nodes, and obtain better results.Since the time from transient node to absorbing nodes relies on the weights on the path and their spatial distance, the background region on the center of image may be salient. We further exploit the equilibrium distribution in an ergodic Markov chain to reduce the absorbed time in the long-range smooth background regions.Furtherly, we cast saliency detection as semi-supervised classification problem in another image graph model. We use the absorbing probability to separate object from background, and reach satisfied results.Finally, we combine robust background prior with the foreground prior obtained from absorbed time in absorbing Markov chain into a unified framework, to optimize the saliency maps and reach better results.Extensive experiments on three benchmark datasets demonstrate the robustness and efficiency of the proposed method against the state-of-the-art methods.