This paper presents a region-based model for salient object detec- tion. The model is defined as a combination of saliency maps con- structed on a hierarchy of image partitions. The saliency map at each level in the hierarchy takes into account the color similarity between each region and its neighbors and highlights objects at the scale de- fined by the partition, with accurate boundaries. Experimental re- sults on a publicly available dataset show that the proposed model outperforms many state-of-the-art saliency methods.

Peer Reviewed