Research On The Approch For Saliency Detection Based On Comparison And Analysis Of Region Feature

The high speed development of information technology has led to the rapidaccumulation of image resources and explosive growth, how to efficiently andintelligently deal with those image data has become the focus of attention. Researcherspropose image salient region detection method to extract the important region of the image,which greatly improves the efficiency of image processing. In order to further improve thedetection effect of salient region, in this paper, image salient region detection methodswhich are based on region are studied.Firstly, two low-level features as well as their common extraction methods weresummarized, and two classic saliency detection models are analyzed.Secondly, salient region detection method base on multi-scale local features andglobal features difference is proposed. It extracted local color feature of each patch afterimage is uniformly divided, then combining the spatial distance to calculate its globalcolor feature differences，and obtaining a preliminary saliency map of a single scale.Finally, the final saliency map with high quality is got by fusing the preliminary saliencymaps of different scales.Thirdly, by further researching characteristics of salient region, a multi-level fusionapproach for detecting salient region in an image is proposed. At first, low-level regioncolor, texture features and spatial coherence in an image are fused to get contrast map ofdifferent region. Then a high-level prior map is generated according to center and colorprior knowledge, then the preliminary saliency map was obtained by fusing the high-levelprior map and region contrast map. At last, the final saliency map with high quality is gotby fusing the preliminary saliency maps of different scales.Finally, experimental results show that the proposed method is feasible and effective,and the analysis of experimental results are made.