Abstract:

To extract important areas in an image that attract visual attention, a saliency detection scheme using both contrast features and local sharpness in the image is proposed. The local sharpness is related to the focusing status, which reflects the human judgment of the scene being imaged. Therefore, the presented method integrates two different visual characteristics, which we call low- and high-level human visual attention (HVA). The globe and local color contrast in the image arouse the low-level HVA, whereas the local sharpness affects the high-level HVA. Two saliency maps are thus generated with color classification and detailed area detection, respectively, and combined. Experiments show that integration of the saliency maps can provide good correspondence to HVA areas for different images.

Key words: visual saliency, color contrast, sharpness region detection