Detecting the Regions of Interest with Edge and Depth Information
Jinye Peng · Jisuanji fangzhen · 2009
Traditional saliency map based regions of interest(ROI) detection shows some difference from the human vision perception,and more than one ROI detection results may appear in the same object during ROI transfer process.In this paper,two kinds of features which are edge and depth information are introduced during feature extraction to access more fitable results to human vision perception,and the dilating process is used for the binary saliency map to connect parts in one object together.Experimental results show a 82.5% tally ratio of the modified ROI detection to human vision perception,which exceeds Itti's 67.5% greatly.