Improved Bayesian Saliency Detection Based on BING and Graph Model

Saliency detection plays an important role in many computer vision applications. The traditional Bayesian based saliency model using convex hull to circle a coarse salient region, which is inaccurate and unstable. To address this problem, we propose an improved Bayesian framework based saliency method. Firstly, we utilize the BING (Binarized Normed Gradients) method to generate the coarse conspicuity map. Then, we construct a graph model after SLIC super- pixel image abstraction, to refine the initial conspicuity map. This is followed by the spatial information based weighting, to produce the final prior map. Secondly, after adaptive threshold, the observation likelihood map is computed by color histogram. Finally, these two maps are combined through Bayesian formula. Experimental results on two benchmark datasets MSRA-1000 and SOD show that our improved method is superior to 13 state-of-the-art alternatives, especially the previous Bayesian saliency models.