Saliency Detection by Superpixel-Based Sparse Representation

Guangyao Chen, Zhenzhong Chen

Abstract: In this paper, we propose a novel model for predicting visual saliency by superpixel-based sparse representation. A superpixel-based sparse representation utilizes the Simultaneous Orthogonal Matching Pursuit algorithm to extract the sparse features from color maps and activation maps of complex cells. The saliency is calculated according to the sparse features from different dictionaries. To guarantee the robustness of the proposed method, the proposed method is performed on a multi-scale basis thus the final saliency result is obtained by using the saliency maps from different scales. Experimental results on multiple datasets show that the proposed model outperforms several advanced methods for saliency prediction.