Stereo Image Quality Evaluation Based On The Correlation Between Multi-resolution And Depth Perception

With the development of information technology,the application of stereo images in modern life is becoming more and more widespread.In the process of stereo image collection,storage,encoding,and transmission,various noises will be introduced to cause image distortion,which will seriously affect the human visual experience and bring bad subjective feelings.Therefore,the research on stereo image quality evaluation technology occupies a vital position in the human application life experience,and has very important practical significance for promoting the popularization and application of stereo image technology.Because the subjective evaluation method is complicated to operate and the cost is too high,how to establish a set of objective quality evaluation models that conform to the human visual system is a hot and difficult point of current research.The human eye is the ultimate receiver of image information,so the human visual perception characteristics play a very important role in establishing an objective evaluation model consistent with the subjective perception of the human eye.The current objective evaluation models for stereo images have problems such as insufficient consideration of human visual perception characteristics,inability to perceive depth information in depth,and algorithms not being universal.Aiming at the above problems,this article will focus on the correlation between the depth information and the visual perception characteristics of the human eye.The main content is divided into two parts:(1)This thesis proposes a binocular fusion stereo image quality evaluation method based on depth perception.Depth information is often represented by disparity maps.At present,research on disparity maps usually incorporates them into brightness images to extract features,however,such processing will lose part of the depth information.Therefore,in this thesis,disparity maps are studied separately,combined with multiresolution models and binocular contrast sensitivity characteristics,to generate multiresolution depth edge sensitivity one-eye maps.This treatment not only considers the physiological mechanism of left and right eye asymmetry,but also reflects the sensitivity of the human eye to the texture of the depth edge at different viewing distances.The combination of multi-resolution and disparity maps not only conforms to the characteristics of human visual perception,but also is not easy to lose depth information.At the same time,the similarity deviation algorithm is used to process the reference and distortion components.This operation can make the extracted features more contrastive.Through comparative analysis of experimental results,it can be found that the proposed algorithm has strong competitive advantages,high ability to fit human visual perception system,and certain generality.(2)This thesis proposes a stereo image quality evaluation method based on multiresolution perception characteristics.This method is based on the subjective perception system of the human eye and changes the processing steps for disparity maps and multiresolution models.Firstly,the image is input into a multi-resolution model,and then the algorithm sequence of the disparity map is calculated to construct the visual perception scene of the human eye at different distances of the viewing object.This processing is more in line with the receiving processing mechanism of human vision.Finally,the gradient cross-mapping bias value is estimated for the multi-resolution disparity map,which makes the extracted depth features more simple and effective.At the same time,in order to more comprehensively extract image features,a hierarchical structure analysis of the brightness image is added.The experimental results show that after fully considering the subjective perceptual characteristics of the human eye,the performance of the algorithm has been significantly improved,which indicates that the objective values predicted by the proposed algorithm are consistent with the subjective values.