Foreground And Background Fusion For Salient Object Detection And Application

With the rapid development of computer technology and the popularization of the Internet,it is a great challenge for people to accurately and quickly extract the most important information from the massive natural images.Nowadays,With the rapid development of computer vision,saliency detection is one of the important branches in the field of computer vision.Salient object detection is able to quickly and accurately identify the most useful and informative salient object regions in natural scenes,which makes great contributions to its related fields.Saliency detection is widely used in computer vision related content such as image segmentation,object recognition,image retrieval and image compression.In this thesis,the propagation algorithm of graph model is used to detect salient objects.Most of the existing ctypical saliency detection algorithms based on graph model use segmentation of background to highlight salient object regions,while in this thesis,candidate object regions are used as the fusion of foreground features and background features for salient object detection.In this thesis,salient object detection and application based on foreground and background fusion are proposed.The realization process of salient object detection algorithm based on foreground and background fusion(Foreground and Background Fusion,FBF)is as follows: firstly,superpixel segmentation algorithm(Simple Linear Iterative Clustering,SLIC)is used to segment the image superpixel.Secondly,the candidate object region is obtained based on the convex-hull model.If the salient object is in the background boundary,the boundary superpixel region will be removed,so that the superpixels around the remaining images will be the background region.The saliency maps of foreground and background features are obtained respectively by using the propagation algorithm of the graph model.Finally,the saliency maps are fused by Bayesian fusion algorithm.The proposed algorithm and the existing about ten typical saliency detection algorithms were tested on several public datasets.The experimental results show that the salient region obtained by the proposed algorithm is more complete and refine the boundary information,The comprehensive results of various evaluation indexes show that the proposed algorithm can achieve good performance.Based on the salient object detection algorithm proposed in this thesis,an image salient object detection system is designed and implemented by MATLAB tool.For any input image,the system can perform image preprocessing,that is,superpixel segmentation,salient object detection to obtain the corresponding salient map,and the salient detection algorithm to evaluate the performance index.