PROJECT TITLE :

Saliency Detection via Absorbing Markov Chain With Learnt Transition Probability - 2018

ABSTRACT:

During this Project, we propose a bottom-up saliency model based on absorbing Markov chain (AMC). Initial, a sparsely connected graph is made to capture the local context data of every node. All image boundary nodes and alternative nodes are, respectively, treated because the absorbing nodes and transient nodes in the absorbing Markov chain. Then, the expected range of times from every transient node to any or all other transient nodes can be used to represent the saliency price of this node. The absorbed time depends on the weights on the path and their spatial coordinates, that are utterly encoded within the transition probability matrix. Considering the importance of this matrix, we tend to adopt completely different hierarchies of deep options extracted from fully convolutional networks and learn a transition chance matrix, which is called learnt transition likelihood matrix. Though the performance is significantly promoted, salient objects don't seem to be uniformly highlighted terribly well. To solve this drawback, an angular embedding technique is investigated to refine the saliency results. Based mostly on pairwise local orderings, which are made by the saliency maps of AMC and boundary maps, we rearrange the worldwide orderings (saliency value) of all nodes. In depth experiments demonstrate that the proposed algorithm outperforms the state-of-the-art methods on six publicly accessible benchmark data sets.