Abstract
Visual attention is usually attracted by the videos which have motion salient regions, so motion is very important to human’s visual attention analysis. In this paper we propose a novel method for motion salient region detection by analyzing region-of-non-interest (RONI) spatial-temporal information. RONI is exploited for the first time in this paper. Different from former methods, our approach starts from the RONI, and looks for motion salient regions as their complements. Position as a prior knowledge is used to determine the initial region-of-non-interest (IRONI). Together with bottom-up feature, more exact RONI can be obtained to detect motion salient regions. Experimental results indicate that our scheme can detect motion salient region effectively especially to the videos with complicated camera motion.