基于BJND和JNDD的立体视频深度感知增强技术综述

Stereoscopic video depth sensation enhancement technology based on BJND and JNDD

人类视觉系统是一个具有多种视觉特性的复杂处理系统,视觉感知是人眼多种视觉特性共同作用的结果。立体视频图像因视差和深度的存在有着与平面图像不同的视觉感受,且深度感与视差有着紧密的关系。恰可察觉差别(Just Noticeable Difference,JND)或者称为恰可察觉失真(Just Noticeable Distortion,JND)是一种基于视觉心理学和生理学的视觉特性,双目恰可察觉差别(Binocular JND,BJND)和恰可察觉深度差别(Just Noticeable Depth Difference,JNDD)影响人们对立体视频的用户体验。在阐述JND基础上,综述人类视觉感知立体视频的BJND、JNDD模型及国内外研究现状,分析视差图质量和深度图信息对立体视频感知影响,提出了基于JNDD的深度图处理方法,寻求人眼对立体视频深度感知增强技术,建立基于BJND和JNDD的立体视频图像质量客观评价方法,为立体视频图像处理及应用提供指导。

Human vision system is a complex processing system with a variety of visual characteristics, the visual perception result from several visual characteristic effects. Stereoscopic video image has distinct in visual perception from 2D image because of the disparity and depth information, and there is a close relationship between the depth sensation and the disparity. Just noticeable difference （JND） is a visual characteristic based on psychology and physiology, the binocular JND （BJND） and just noticeable depth difference （JNDD） can affect users＇ experiences of the stereoscopic video. Based on introducing JND, this paper summerizes BJND and JNDD models of HVS as well as current research in the international field. By discussing the influences of disparity map quality and depth map information on stereoscopic video perception to improve the stereoscopic video sensation enhancement, a new method for the depth map processing is proposed. The stereoscopic video objective quality assessment model is established based on BJND and JNDD, thus providing the theoretical guidance to the processing and application of the stereoscopic video.