View synthesis distortion elimination filter for depth video coding in 3D video broadcasting

Abstract

Depth image based rendering (DIBR), which can generate synthesized images according to the users' demand, is a key technique for achieving 3D television. However, view synthesis by DIBR technique is very sensitive to depth coding distortion. Because depth distortion will lead to geometrical rendering position errors, and seriously affect the quality of synthesized images. In this paper, we propose an in-loop filter to minimize view synthesis distortion at the cost of transmitting extra filter parameters as supplementary information. And an adaptive parameter determination scheme is presented for the proposed filter. Then a good trade-off between bit rate and view synthesis distortion has been achieved by considering the spatial-temporal correlations of 3D video sequence. The simulation results reveal that the proposed view synthesis distortion elimination method can significantly improve the rate-distortion performance, which achieves Bjontegaard Delta Peak Signal-to-Noise Ratio (BDPSNR) gain from 0.41 to 1.09 dB compared with the benchmark.