文摘
As an important problem in computer vision, saliency detection is essential for image segmentation, super-resolution, object recognition, etc. In this paper, we propose a novel method for saliency detection on image using region contrast and machine learning approaches. An image boundary extension-based general framework is proposed that can be used for all rarity- or sparsity-based schemes to improve their performances. Then, a saliency map based on boundary extension and region contrast is constructed. Due to its unsatisfactory performance, another saliency map combining supervised locality-preserving projection and support vector regression is built, to complement the previous saliency map. A final saliency map can be obtained by fusing these two saliency maps. The proposed method is evaluated on the publicly available dataset MSRA-1000 and compared with 13 state-of-the-art methods. Experimental results indicate that the proposed method outperforms existing schemes both in qualitative and quantitative comparisons.