A virtual view distortion estimation model for depth map coding

The 3Dvideo format of multi-view video plus depth(MVD)is a research hotspot nowadays.The depth map provides the scene geometry information and does not show up in the display side,but is used to render the virtual view at the terminal by using the depth image based rendering(DIBR)technique.The characteristics of the depth map used for rendering instead of displaying means that accurate estimation of the virtual view distortion can improve the rate distortion performance of depth map coding.The procedure of depth map compression may cause distortion,and the depth map distortion will cause the distortion of the rendered virtual view.This paper analyzes different virtual view distortions induced by different depth map distortions and proposes an exponential model to estimate the virtual view distortion induced by the depth map distortion,and further employs the model in the rate distortion optimization(RDO)for depth map coding.Experimental results show that the proposed model can estimate the virtual view distortion induced by the the depth map distortion accurately and improve depth map coding performance.Compared with the VSO in HTM,the proposed model reduces the encoding time for about 10%and the virtual view quality is even a bit better than that of HTM.