(A) novel error concealment method based on DCT-coefficient grouping and interleaving in wireless video coding

Abstract

In this thesis, we solve the problem of burst error in video transmission through a wireless channel. Burst error damages the information so severely that it is very hard for the decoder to recover the lost information. Dispersing the burst error makes a good situation for the decoder to recover the damaged information, because it has more information to use in the restoration process than without spreading out the burst error. In this thesis, we propose a coefficient grouping method to spread out the burst error. Transform coefficients are grouped into several coefficient groups (CGs), and each CG is encoded in the conventional encoding procedure. The CGs are grouped in a way that each CG has as equal importance as possible. The generated bits from each CG are interleaved in a bit stream so that burst error can not destroy all CGs from one block. Error resilient entropy coding (EREC) is adopted to make the bit stream resilient against the error propagation through the CG boundary. We also propose new error concealment algorithms based on projection onto convex sets (POCS) which use the clean information which is not damaged by burst error. The correctly received transform coefficients thanks to the coefficient grouping and interleaving (CGI) method are utilized to make an additional convex set. It is pointed out that this simple application of POCS to the CGI method is not enough to conceal the damaged block to a satisfactory degree. We note that POCS has a strong initial value dependency. The part of correctly received coefficients is again utilized to find out a proper initial value for POCS iteration. The simulation results show that the proposed concealment method in conjunction with the CGI scheme performs well in restoring the missing information.