Distributed video coding with limited feedback requests

Abstract

Traditional decoder rate control (DRC) distributed video coding (DVC) can dynamically adjust the amount of parity information through feedback channel, which guarantees that minimum parity information should be transmitted from the encoder to the decoder. However, frequent feedback requests increase system complexity and video transmission delay. To solve the problems caused by frequent requests, this paper proposes a new feedback control algorithm which purposefully requests for more parity information through the feedback channel. To achieve a good rate distortion (RD), the proposed DVC architecture also performs frame classification, side information refinement and correlation noise model refinement. Frame classification module makes the judgment whether the current frame worth coding according to the motion intensity, which reduces the complexity of DVC system. Side information and correlation noise model refinement are performed based on correctly decoded bit-planes, which contributes to decreasing the error rate of the decoded frame. Experimental results show that the new feedback control algorithm significantly decreases the number of feedback requests. A good rate RD performance is also achieved with only one request. The proposed architecture has a better RD performance than encoder rate control (ERC) systems, and it also achieves a similar performance to DRC systems without feedback constraints in literature.