IEEE Transactions on Consumer Electronics, Vol. 54, No. 4, NOVEMBER 2008
Contributed Paper
Manuscript received August 19, 2008 0098 3063/08/$20.00 © 2008 IEEE
1846
Adaptive Temporal Error Concealment Scheme
for H.264/AVC Video Decoder
Yanling Xu and Yuanhua Zhou, Member, IEEE

Abstract — Resilience to channel noise is indispensable
for video communication based consumer applications.
Temporal error concealment is a kind of decoder-based
technique to compensate for transmission errors and is
popular in most consumer electronics. In this paper, an
adaptive temporal error concealment scheme is proposed
based on H.264 video standard to improve error resilience
ability for video consumer applications. Mode switching
mechanism is devised to flexibly choose appropriate temporal
concealment mode. The search range for candidate motion
vectors could be adaptively determined by the range of
spatially and temporally neighboring motion vectors.
Weighted outer boundary distortion function could
dynamically select the optimal replacing motion vector for
local area. Experiment and comparison with recent error
concealment schemes shows that our proposed algorithm
could effectively improve received video quality with reduced
requirement for decoding time1.

Index Terms — Error concealment, video, temporal,
motion vector.
I. INTRODUCTION
Video consumer electronics, such as mobile TV, 3G
cellular phones and portable video terminals, have become
more and more popular in mobile and/or wireless applications.
The latest standard H.264/AVC[1], the most promising video
compression standard for a large range of video consumer
applications, adopted many new strategies to greatly improve
video coding efficiency, such as Intra prediction, tree
structured motion estimation and flexible block sizes.
However, the transmission of compressed video streams
over error-prone networks is usually vulnerable to channel
disturbances such as bit errors and packet losses, which leads
to data corruption and serious video quality deterioration.
Hence, error resilient techniques are highly desirable for video
transmission through unreliable channels.
Error concealment (EC) is involved in most portable
consumer electronics to enhance the reconstructed video
performance. As the post-processing technique implemented
at the decoder side, EC has the attractive advantage that it is
encoder-independent and has no requirement for modification

1 This work was supported by the National High Technology Research and
Development Program of P. R. China (863 Program).
Yanling Xu is with Tongji University, Shanghai, P.R.China (e-mail:
xuyanling@mail.tongji.edu.cn).
Yuanhua Zhou is with Department of Electronics Engineering, Shanghai
Jiaotong University, Shanghai, P.R.China.
of video standards. It intends to compensate the effects of
transmission errors by exploiting correlated information in the
video signal to provide an approximation of the lost data.
Depending on the information used, error concealment can be
divided into two categories, namely, spatial error concealment
(SEC) and temporal error concealment (TEC). The former
exploits the spatial redundancy in the same frame, while the
latter utilizes temporal redundancy between successive
frames.
The reference model of H.264/AVC provides temporal
error concealment for Inter-frames on the level of 16×16
macroblock (MB)[2]. Motion vector (MV) of the corrupted
macroblock is estimated from motion information of the four
spatial neighboring macroblocks. However, this scheme has
low capability in restoring detail movements. It performs well
only when the damaged macroblock lies in a smooth region
with low motion. Otherwise, it results in severe blocking
artifacts or fragmentation of edges.
Other TEC methods have been reported focusing on the
resumption of lost MVs. The MV candidate set[3] may consist
of available neighboring MVs and their average or median
MVs. Boundary Matching Algorithm (BMA) is presented in
[4] [5] to select the replacing MV. It is refined in [6] with
weighted interpolation. [4] takes into consideration the
boundary distortion of macroblocks affected by error
propagation in future frames, which will introduce additional
decoding delay. [7] proposes the overlapped motion
compensation method in which three subblocks predicted
from the side match criterion and the MVs of vertically and
horizontally adjacent subblocks are averaged and merged
using a weighting matrix. [8] provides the Lagrange
interpolation formula to describe the motion tendency of local
area and to estimate the lost MV. [9] proposes a customized
polynomial interpolation model built with the adjacent MVs to
recover the missing MV. Most of these methods improve the
performance at the cost of higher computational complexity.
This paper aims to propose an efficient temporal error
concealment scheme at a reduced complexity. The
concealment mode can be switched flexibly according to local
motion conditions employing some new characters of H.264.
The set of replacing motion vector candidates is dynamically
formulated based on the distributing range of neighboring
vectors. The matching criterion for the optimal motion vector
is enhanced with adaptive outer boundary matching measure.
The rest of the paper is organized as follows. Section II
describes the proposed temporal error concealment from three
aspects. In section III, simulation experiments are performed
and results are discussed. Section IV draws the conclusions.