Closed Salient Edge Extraction Based on Visual Attention Mechanism

Abstract

Edge detection is a means of generating compact descriptions which preserve most of the structural information in an image. However, the edge image returned by the edge operators usually contains false edge points or exhibits gaps in edges which are generated by image noise. To solve the problem, a closed salient edge extraction approach was proposed, which makes use of the visual attention model given out by Itti together with the original edge image to obtain the salient edges and then the salient-edge gaps were linked so that the closed salient edge of an image can be extracted. In order to link the edges, a membership function based on the gradient and directions of endpoints of the salient edge image was carried out to control the extension of endpoints. Experimental results showed that the proposed algorithm was able to extract the closed salient edge of an image effectively especially for the image whose background was not very complex.