Error concealment plays an important role in video streaming applications under error-prone environments. This scheme tries to minimize the visual artifacts due to errors which include bit conversions/drops and packet loss. Since error concealment technique is non-normative in H.264/AVC standard, there are possibilities that any appropriate algorithm can be incorporated into the standard. The decoder utilizes the spatial, spectral, and/or temporal redundancies of the decoded video data to perform the error concealment.
In this thesis, we propose temporal error concealment methods for H.264/AVC. Depending on the Flexible Macroblock Ordering (FMO) types, we develop different algorithms to perform the temporal error concealment. With the type 0 of FMO, interleaving type, based on the reference frames of the neighboring macroblocks, the algorithms adaptively perform extrapolation to recover the motion vectors of lost macroblocks with a boundary matching algorithm. With the type 1 of FMO, check-board type, based on the mode of four neighboring macroblocks, the mode of a lost macroblock is predicted. With the given mode, using the motion vector of neighboring blocks, all the sub-blocks of the lost macroblock are concealed with the estimated block pixels.
To assess the proposed error concealment methods, the experiments were conducted for various types of test video sequences. The experimental results show that the proposed methods outperform the conventional methods in terms of image quality with 3.74 dB in maximum.
