Error Resilience Analysis of Motion Vector Prediction in HEVC
Carreira, J. F. M.
; Ekmekcioglu, E.E.
; Kondoz, A.
Error Resilience Analysis of Motion Vector Prediction in HEVC, Proc Conf. on Telecommunications - ConfTele, Aveiro, Portugal, Vol. 0, pp. 1 - 4, September, 2015.
Digital Object Identifier:
Download Full text PDF ( 184 KBs)
The High Efficiency Video Coding standard achieves a high level of compression efficiency when compared to previous standards. As a counter effect, its error resilience is decreased and performance under error prone conditions is greatly affected. In this paper the error resilience of HEVC is evaluated, focusing on the new motion vector prediction technique referred to as Merge Mode, by analysing the impact of using different number of motion vector (MV) candidates and temporal MV dependencies. Since spatial dependencies are not the most relevant for error resilience, the focus of this performance study is on the temporal dependency of motion information due to its greater impact on error propagation. The results indicate that quality gains can be achieved by disabling the temporal motion vector candidate (TMVP), without compromising the coding efficiency. In our experiments average quality gains up to 5.17 dB (PSNR) can be achieved when TMVP is disabled for a 10% of loss ratio.