Allowable depth distortion based fast mode decision and reference frame selection for 3D depth coding.

To improve the coding efficiency and meanwhile reduce the complexity of the depth video encoder in 3D system, we present an efficient depth coding scheme based on the allowable depth distortion (ADD) in view synthesis. Firstly, we analyze the depth distortion in view synthesis and present a new ADD-based Rate Distortion (RD) optimization cost function to improve the mode/reference decision criteria. Then, we present an early mode and reference frame selection algorithm, which skips the unnecessary and complex mode and reference frame when the depth distortion is within the ADD interval. Meanwhile, the coding efficiency is improved by RD optimized mode decision and reference frame selection. The experimental results show that the proposed overall algorithm, which integrates the fast mode decision, reference frame selection and RD optimization, can reduce the depth coding complexity from 32.47 to 72.59, and 55.95 % on average over different test sequences. Meanwhile, the proposed algorithm achieves 1.34 and 1.58 dB average Bjontegaard Delta peak signal-to-noise ratio gain in terms of the view synthesis quality at low and high bit rate, respectively. In terms of the bit rate reduction, it reduces 54.66 and 49.60 % Bjontegaard Delta bit rate on average at low and high bit rate, respectively, while maintaining the same view synthesis image quality. [ABSTRACT FROM AUTHOR]