Research On Image Salient Region Detaction Algorithm

With the proposal of the "Internet plus" national strategic plan,it is to Internet drive newly,more and more information is transmitted by images,extracting valuable information quickly and efficiently in mass image data is the hot issues in computer vision currently.Saliency detection algorithm is affected by the selection of human visual attention,and simulate the mechanism to catch the noticeable areas in the image,and to ignore the information in which there is no content,thus the efficiency of image processing is improved.So image saliency detection algorithms have applied into many fields,such as object detection and recognition,image retrieval,image compression and target tracking in video.There are a lot of research on image saliency detection,by analyzing the shortcomings of existing saliency detection methods,this paper combined with biological visual features and image multi features and superpixel segmentation proposed an improved saliency detection algorithm.The work of this paper includes the following aspects:(1)The algorithm is based on bottom-up approach to detect the salient area.Firstly,we describe the the selection of human visual attention mechanism,and introduce the basic principles of the bottom-up visual analysis and computation algorithms in detail,and then enumerate several typical saliency detection algorithms and superpixel segmentation method,in which the characteristics of them are expounded and analyzed in detail.(2)We proposed a detection method based on multi-scale and multi-feature for saliency detection.Ours method is based on the FT algorithm,and used the improved eight neighborhood method and the entropy-based superpixel segmentation method to obtain the brightness characteristics,color features and texture features of the images at various scales so that the contours of the significant regions are more obvious.In order to verify the performance of the algorithm,the algorithm was tested on five public datasets,compared with twelve existing algorithms for saliency map and ROC curve,accuracy rate and recall rate.The experimental results show that the proposed method can improve the detection effect of the image effectively.(3)We analytic study the extracting of seal in Chinese painting and calligraphy works based on saliency detection.Using the framework of multi-scale and multi-feature saliency detection method to obtain the brightness characteristics,color features and priori features of the images at various scales,aiming at the characteristics of seal area in painting and calligraphy works,select red prior feature and shape prior feature to constrain,then identify the red seal area in the image,and it achieved good experimental results.