Region Detection And Application Based On Visual Saliency

Salient region detection has shown more and more broad prospects in the field of image processing. The high quality and reliable saliency map eliminates the redundant information and improves the exclusive characteristics of image and the effect of second processing to a certain extent, but how effectively and robustly obtain a salient region of image is still concerned. We mainly completed the following works around the theory and application research in visual saliency detection in this paper:(1) The paper starts from the theories and techniques of image processing, focused on the research of the mechanism and the existing visual salient region detection model of human visual attention, and analyzed several typical algorithms and applications.(2) Salient detection method based on super pixel contrast was presented on the base of the context of contrast detection method, which is different from previous methods based on the contrast.First, the image divided into some region in which the pixel has the similar characteristics and size by using Super pixels segmentation.Computer the saliency of super pixel as the saliency of image by using the method based on the context of contrast, A method based on double super pixel fusion was proposed in the case of analysing the result of the super-pixel detection in coarse segmentation and fine segmentation, Last, A saliency map with the full resolution was got by using the method of adaptive threshold to enhance the regional saliency and reduce the saliency of background region.(3)In order to verify the effective and feasibility of the method in the field of image saliency detection, it was made experiments on one of the largest publicly available data sets,We apply this method on the salient ob ject extraction and image retrieval. We make difference by comparing this method with several others saliency detection method using several common evaluation method. The experimental results show that the method proposed in this paper is effective and feasible in terms of saliency detection.