Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4970526 | Signal Processing: Image Communication | 2016 | 24 Pages |
Abstract
Saliency detection is a challenging issue in computer vision, but one of great importance and huge applications. In this paper, we propose a novel Graph-Boolean Map method to automatically detect salient regions. First, a few Boolean maps are generated based on multiple channels by applying gamma values. Second, we perform a multi-scale propagation method by learning a new weight matrix based on two important priors, and encourage superpixels with similar appearances share similar saliency scores, yielding more stable results. Next, we generate the final saliency map by applying a graph inference based on belief propagation. Experimental results demonstrate that the proposed method significantly outperforms several state-of-the-art methods to salient object detection. Our approach is also applied for infrared images.
Keywords
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Vision and Pattern Recognition
Authors
Wei Qi, Jing Han, Yi Zhang, Lian-fa Bai,