New just-Noticeable-Distortion Model Based on the Depth Information and its Application in Multiview Video Coding

Abstract

To reach higher compression ratios without perceptually degrading the reconstructed signal, the properties of the human visual system (HVS) need to be better exploited. Research simulated the sensitivity of HVS with the just-noticeable-distortion (JND) model, which describes the perception redundancy quantitatively. This paper proposed a new JND model in multiview video coding (MVC) which explored the property of HVS to stereoscopic masking effect. The proposed model contained the spatial JND, temporal JND and the JND in depth. A novel asymmetric stereoscopic video coding method was proposed based on the new JND model. For each macroblock (MB), the quantization parameters (QPs) were optimized adaptively based on the new JND. Compared with the original JMVC scheme without JND, our method gets remain perceptual quality and save bits averagely 24.18%~36.24%.