A model of visual attention for locating region of interest in large background

A computational model of visual attention was developed for regions of small defects in a large background for inspection of surface quality in the line, simulating human visual attention mechanism. Firstly global features were extracted from an input image by Law´s rules. Secondly Local features were extracted and evaluated through Itti´s method and inspection requirements. Then the local features were fused a saliency map by a new fusion operator which are not the same with Itti model. Finally the region of interest is “pop-out” in the map. Experimental results show that the model can locate the region of interest, and remove the most background regions.