Salient Region Detection And Its Application Based On Multi-scale And Multi-feature

With the rapid development of computer and Internet technology,large amount of data increase every day,including the digital image resources.For the massive data processing of pictures and videos,it is becoming more and more important.How to efficiently receive,analyze and process these image information and extract the interested parts quickly is an important research topic.Through the visual system,humans can quickly search for targets of interest in complex scenes,and the computer simulates the human visual system to obtain salient regions in the image,that is,salient regions detection.The priority processing region is obtained by the saliency region detection,so that limited computing resources are allocated to important information,and the calculation efficiency is improved.It has important application value in target recognition,image compression,image retrieval and image classification,and it is also a hot topic in current research.Although a variety of salient detection methods have been proposed at present,there are still a series of challenges,and the quality of the image saliency detection results is not high,and the salient regions cannot be completely segmented.The traditional saliency detection method does not consistently highlight the entire object for some complex background images,and cannot clearly indicate the outline of the object.Exiating these problems,two salient regional detection algorithms are proposed in this paper.One of them is Bayesian model saliency detection algorithm based on multiple scales and improved convex hull.Another is Salient region detection based on improved manifold ranking.The main research work of this paper includes:1.Traditional Bayesian model saliency detection algorithm may have a poor performance in terms of precision..Therefore,a novel algorithm is proposed in this paper,which is based on the multi-scaled convex hull.Firstly,the Manifold Ranking(MR)algorithm is used to extract the foreground of the images in the CIELab color space,which is considered as the prior probability map.Secondly,the image is down sampled by Gaussian Pyramid algorithm and,results in three scaled images.The improved convex hull is derived by using the intersection about convex hull of Harris corners of the three scaled images.Thirdly,the color histogram and convex hull are combined to calculate the observation likelihood probability.Finally,according to the existing prior probability map and observation likelihood probability,the Bayesian model is used to compute the saliency map.2.Existing graph-based manifold ranking based salient object detection algorithm is less effective in detecting images with complex background.This paper proposes an improved algorithm based on Multiple Kernel Boosting(MKB)method and manifold ranking.Firstly,the image is divided into four different super-pixel scales.Then according to the RGB color space feature,the CIELab color space feature and the LBP features,calculated the boundary saliency maps of the upper,lower,left and right directions of the four scale images.The saliency maps of four directions are combined to obtain the four-scale image saliency map,respectively.The weak saliency map is computed by the saliency map of the four-scale image.Secondly,generated a training sample based on the weak saliency map.A strong classifier based on samples directly from an input image is learned,in order to detect salient pixels by the multicore enhancement(MKB)algorithm.Finally,the integrated multi-scale saliency map further improves the detection performance.In addition,optimized the process to get the final saliency map.3.Moreover,the optimization is carried out for better performance.In this paper,the proposed algorithm is analyzed experimentally on three standard data sets of MSRA1000,ECSSD and PASCAL-S.Compare the two saliency algorithms proposed in this paper with 12 widely used saliency algorithms.The saliency algorithm has a good detection effect in both simple scenes and complex backgrounds.The detection accuracy is better than the above 12 saliency algorithms.Meanwhile,the experimental time comparison of various algorithm time complexity.Our algorithm not only achieve good vision effect,but also improves the performance evaluation of precision-recall curves and F-measure values.Finally,the saliency algorithm of this paper is applied to the actual scene of traffic sign detection.Experiments show that the algorithm can detect complete traffic signs quickly and accurately.