Studies On Saliency Detection For Stereoscopic Images

The technology of saliency detection refers to using computers to simulate the visual attention mechanism of the human eye,and extract the target areas of human interest from the images quickly.Reliable saliency detection results can be widely used in applications such as target recognition,anomaly detection,image retrieval and image classification.Therefore,saliency detection has become a hot topic for many scholars.Most of the existing saliency detection methods are using two dimensional features such as color,texture,and brightness.With the development of binocular stereo cameras and Kinect cameras,the acquisition of depth information has become very easy.The stereoscopic images with depth information are closer to the biological vision system.The depth information provided by the stereoscopic image can effectively solve the problems of the saliency detection results based on the two dimensional images,such as light noise,similar foreground background and multi-target occlusion.In this paper,we study the saliency detection technology for stereoscopic images using the color and depth information of the scene.First,the characteristics of depth information are analyzed.And a saliency detection method based on depth selective difference is given.At the same time,analyzing the distribution characteristics of background information in the image,an improved saliency detection method based on background prior is given.Next,the above two kinds of saliency detection methods are merged,and optimization measures are added to obtain the final saliency map based on depth selective difference and background prior.Then,the saliency map is used in object segmentation and image scaling.Finally,the saliency detection and application system for stereoscopic images is given.The main work of this paper is as follows:Firstly,a new saliency detection method based on depth selective difference and background prior is given.Aiming at the characteristics of depth information in the image and the distribution characteristics of the background information,an image saliency detection method based on depth selective difference and background prior is proposed.Finally,qualitative and quantitative experimental analysis on two published datasets proves the validity of the saliency detection method.Secondly,two specific applications based on saliency detection are implemented.Based on the experimental results of saliency detection based on depth selective difference and background prior,two applications of object segmentation and image scaling based on saliency detection are implemented.Through experiments,it proves the superiority and effectiveness of the proposed saliency detection method in these applications.Finally,we design and implement a saliency detection and application system for stereoscopic images.We introduce the system design of the system and present the system function at first.Then the experimental results of the system are evaluated.