Target Recognition Algorithm Based On Local Spatial Redundancy Suppression Of Visual Information

With the development of the computer,the application of the image has been heavily into every aspect of our life,more and more images need us to identify.Therefore,in the field of image processing,target recognition has become a very important part,and has a strict requirement for the reliability of the results.In the face of the background of the input image is very complex and contains a large amount of redundant information,the region of interest(ROI),a partial region can cause visual attention,is possible to quickly attract the attention of the human visual system and makes the human visual system to prioritize and final recognition,thereby ignoring the redundant information,namely those not cause the attention of the visual information.This rapid concentration in the area of interest,to get the main information of mechanism in the input image is referred to as visual attention mechanism of human visual system.To the image of interested area is extracted by processing,can not only speed up the speed of image processing,but also can improve the accuracy of target recognition.In view of the excellent ability of human visual system to process the image information,in this paper,the visual attention mechanism is applied to the target recognition of the image.A new method to recognize target of visual attention model based on superpixels and redundancy reduction is proposed,it is about the target recognition problem under the similar redundancy background.Firstly,in order to get similar rigions,the image is segmented based on optimized spatial feature of superpixels.Secondly,to eliminate the correlation between pixels,the Shannon entropy is considered to represent pixels information which build up the image information map.Finally,aiming at suppressing the repeated items effectively,the redundancy of the information map is thrown away to yield the image saliency map using the self-similarity.Experimental results show that the proposed algorithm is comparable with the state-of-the-art saliency models,which not only highlights the salient objects in a complex environment but also more effectively suppress the redundancy in the background of the image and improve the recognition efficiency.