An efficient request stopping method at the turbo decoder in distributed video coding

Most of the literature on distributed video coding (DVC) is based on a Wyner-Ziv video coding architecture that adopts turbo codes to replace conventional source coding tools. By moving the motion estimation task at the decoder, it achieves light encoding complexity. Nevertheless, this scheme has a serious limitation that hinders its practical application. In fact, the decoder needs an efficient way to estimate the probability of error without assuming the availability of the original video at the decoder. The original frames are used to detect when a sufficient number of parity bits are received, in order to guarantee a residual BER (bit-error rate) below a given threshold, typically set to be equal to 10
^-3
. In this paper we investigate the use of a practical stopping criterion that can be used to detect successful decoding, without the need to have access to the original frames. We study the robustness of this method for binary i.i.d. sequences transmitted over a BSC channel. We show that there is a tradeoff between residual BER and rate overhead, that can be adjusted by tuning the threshold of the stopping criterion. We apply these concepts to both pixel-domain Wyner-Ziv coding (PDWZ) and transform-domain Wyner-Ziv coding (TDWZ), showing that the rate-distortion performance is only marginally affected by the lack of original frames at the decoder. A maximum loss of 0.1dB is observed at high bit-rates.