Research On3D Video Coding Based On JND

JND is Just Noticeable Distortion. It is the distortion threshold of an area based on theory of human visual system. Because of many nonlinear effects, these thresholds of different areas are different. When the distortion is below this threshold, human eyes should not perceive the distortion. Generally, the goal of perceptual video coding is to enhance the perceptual quality of the video, or to assessment a video based on the perceptual characteristics. And JND measure is an important part of perceptual video coding.H.265/HEVC is the latest video encoding standard based on the hybrid video encoding framework. It is released in2013by ITU-T and ISO/IEC. The3D extension of HEVC is not finalized yet. HEVC3D is not like the3D extension MVC of H.264. For multi-view3D display, HEVC3D encodes a depth map for each view, and apply the virtual viewpoint generation algorithm. This paper analyzes the new features of HEVC3D and the relationship between depth map and disparity value. Based on the human visual attribute on different depths, this paper proposed a JND model using depth information. HEVC3D extension uses the same motion estimation process of HEVC. Motion estimation process searches the best motion vector. This motion vector contains motion information from different area of the image. Human eyes are more concerned on the moving object, but the distortion sensitivity will decrease if the moving speed is too high. Based on the attribute, this paper proposed a JND model based on motion information. This paper using the depth and motion JND as a factor, multiply the spatial JND, get a JND model for the3D video.This paper analyzes the quantization and transformation parts of HEVC3D, and integrate the3D JND model into the encoder. Human eyes are not sensitive to the high distortion threshold areas. If the JND value of one block is higher than the average value, encoder can have more distortion on these areas in the texture image. This paper modifies the encoder and test the encoding. The result shows this method can effectively reduce the bitrate of the video.