Research On Adaptive Feedback Channel Distributed Video Coding Technology

With the rapid development of wireless sensor, mobile telephone, digital camera, PDA and other wireless mobile devices in human’s daily life, Distributed video Coding (DVC), a kind of video compressing method which can adapt to the resource-constrained devices, begins to attract wide attention. DVC moves motion estimation and compensation to decoder, therefore the complexity of coder is inferior to decoder. Meanwhile, having adopted joint source-channel coding, DVC has great error resilience performance which can be applied to wireless video application scene. As DVC scheme based on ideal feedback retransmission strategy is not feasible, this dissertation investigates DVC on ideal feedback channel and no feedback channel. Derived from the above research foundation, this dissertation proposes a self-adapted feedback channel DVC scheme with rate-distortion driven decoder-side.The specific work and innovations of this dissertation can be summarized as follows:1. In the mainstream DVC systems at present, the coding of key frame often partly uses or even adopts some traditional referred technologies about image and video coding technology, such as Intra coding in H.264/AVC. Although using intra frame as key frame can ensure low coding complexity, this method brings low code efficiency compared to current hybird video coding. Based on the above analysis, this dissertation proposes a DVC video coding scheme based on improved feedback channel key frames, mainly considering the method to improve coding efficiency of DVC as much as possible by using feedback channel. Combining with complexity rate distortion model, this dissertation increases DVC overall performance from two aspects. One aspect is low-complexity frame/block motion activity based key frame coding, the other is adaptive3DRS motion estimation algorithm. 2. Typical DVC sehemes are based on feedback channel and decoder controls the distribution of the code rate. This kind of scheme can reach the optimal code rate distribution and the best rate-distortion performance. However, these schemes are not suitable for the application context feedback-free channel or with strict delay requirements. In view of the above problems, this dissertation investigates off-line correlation noise model (CNM) applied to non-feedback DVC. Then a low delay feedback-free DVC frameworks is introduced. We also propose rate estimation algorithm based on off-line CNM and extrapolated side information generating algorithm based on motion compensation for DVC.3. We propose a kind of adaptive DVC method-----a decision mechanism applied to decoder rate-distortion driving mode. This mechanism utilized decoded video frame to produce rate-distortion cost function of the frame, and judges coding mode of DCT transform domain coefficient of the frame. If the mode is Skip, then give up and does not code or transfer. If the mode is WZ encode mode, then transfer using LDPC coding. The coding and decoding framework through this decision mechanism can reduce transferring code rate without affecting video frame peak signal-to-noise ratio. Derived from the above research foundation, this dissertation proposes a adaptive feedback channel DVC scheme with rate-distortion driven decoder-side. Experimental results show that our coding scheme can achieve significant gain compared to state-of-the-art feedback-free DVC codec while still low encoding complexity.