| Distributed video coding (DVC) is a new technology for video compression, which is based on Slepian-Wolf lossless and Wyner-Ziv lossy source coding theory. In contrast with the traditional Hybrid video compression (HVC) system, DVC achieves low-complexity video encoder by transferring the large number of computational complexity induced by motion estimation from the encoder to the decoder, and the correlation between the encoded data is exploited only in the decoder. DVC is used to implement various "uplink" video applications in resource-constrained mobile devices. Therefore, research on DVC has great theoretical significance and practical value. However, typical DVC schemes are almost based on the feedback channel between the encoder and the decoder, and the rate allocation (RA) is controlled by the decoder. Although these schemes can achieve the optimal RA and best rate-distortion (RD) performance, they are not suitable for feedback-free or tight delay constraint video applications. In order to resolve the mentioned problems, the rate allocation and data transmission strategies for feedback-free DVC are studied.A new rate allocation algorithm is proposed for the designed feedback-free transform domain distributed video coding (FFTD-DVC). The proposed algorithm first chooses the best coding mode in terms of the video content, and then predicts the optimal rate using transform domain correlation estimation model. Finally, the proposed algorithm is integrated into the FFTD-DVC system. The experimental results show that the proposed algorithm can achieve the optimal rate allocation, while still maintaining consistent coding efficiency. Compared to the feedback-based rate allocation methods, the proposed algorithm only results in 0.3dB loss in RD performance.In order to resolve the problems of packet loss and bit error caused by the dynamic and unpredictable nature of the wireless channel, and because the motion regions have a very significant impact on the video quality, a motion-based FFTD-DVC algorithm is proposed for error-resilient video transmission. The proposed algorithm first detects motion regions based on the recognition of import motion vectors, and then compresses them with the proposed FFTD-DVC. The decoder receives the syndrome bits and recovers the lossy parts of the motion regions. In the end, the entire video frames are reconstructed. Experimental results show that compared to FEC methods and decoder-based error concealment methods, the proposed algorithm provides improved error-resilience, and improves above 1.7dB RD performance.In order to achieve the 1080p HD video broadcasting, and also taking into account the current 1080i HD video broadcasting, a new algorithm based on the FFTD-DVC is proposed for full HD video broadcasting. The proposed algorithm first compress the skipped field sequences by FFTD-DVC, and generates the syndrome bits. The decoder makes use of the syndrome bits to recover the estimated skipped field sequences, which are generated by the proposed adaptive de-interlacing algorithm. Finally, the progressive video sequence is reconstructed. Experimental results show that, compared to unicast-based and scalable coding-based methods, the proposed algorithm has a better RD performance, which is improved nearly 1dB.The proposed FFTD-DVC scheme is different from the feedback channel based DVC scheme. The rate allocation is controlled by the encoder in the FFTD-DVC scheme, which does not depend on the feedback channel. The advantages are that the delay is low and the complexity of the decoder is relatively low. Based on the proposed FFTD-DVC scheme, a new error-resilient wireless video transmission scheme is proposed to improve the error-resilience and a new full HD video broadcasting scheme is designed to supports both 1080i and 1080p formats.
|
PDF Full Text Request