Stereo Image Quality Evaluation Based On Visual Characteristics And NSS(Natural Scene Statistical Characteristics)

In recent years,stereoscopic image technology has developed rapidly.They can not only create a sense of stereoscopic,but also bring a real sense of presence to the viewers,so more and more scholars have studied and paid attention to it.The quality of stereoscopic image directly affects the viewer's viewing experience.How to evaluate the quality of stereoscopic image effectively has become an important topic in the development of three-dimensional technology.Based on the related research of stereoscopic technology,this paper aims to design an evaluation method that not only conforms to the human stereo perception system,but also can effectively reflect the quality of three-dimensional images.This paper mainly considers the evaluation of the quality of stereoscopic images from two aspects.On the one hand,it is the content quality that can reflect the degree of image distortion.In this paper,we use DCT transform and wavelet transform technology to analyze the statistical characteristics of original and distorted images.On the other hand,it is based on the stereoscopic perception of binocular vision perception,depth perception and so on.Since human beings are the ultimate recipients of stereoscopic images,the evaluation of stereo image quality should be based on stereo perception system,which is consistent with the characteristics of human perception of stereoscopic images.In this paper,based on the parallax theory,a depth perception map that can directly reflect the human stereoscopic sense is established.Based on this consideration,this paper take the left and right views as reference images and filter the left and right views with the interactive filter,combining summation and difference theory to establish interactive filtering and map based on binocular difference.This paper experiments on LIVE and MCL stereo image database.The results show that the stereo image quality evaluation method based on stereo perception and natural scene statistics is highly consistent with subjective evaluation method.