分布式视频编码的关键技术研究

项目名称： 分布式视频编码的关键技术研究

项目编号： No.60802032

项目类型： 青年科学基金项目

立项/批准年度： 2009

项目学科： 无线电电子学、电信技术

项目作者： 宋彬

作者单位： 西安电子科技大学

项目金额： 20万元

中文摘要： 由于其非对称性，终端简单性和鲁棒性，分布式编码在无线通信中有着广泛的潜在应用。本项目基于自主实现的无反馈分布式视频编码(UDVC)系统，主要研究内容包括：基于原始图像相关性的边信息生成算法；自适应相关噪声模型的构造算法；Wyner-Ziv帧DC（直流）系数的准确率失真模型构造；无反馈分布式视频编码的编码端速率控制算法；具有鲁棒性的Wyner-Ziv帧重构算法；保证Wyner-Ziv帧质量的抗误码方法；分布式视频压缩感知算法等。将上述多种关键技术应用在UDVC系统上，实验结果表明，UDVC编码端复杂度小于H.263编码器的20%；与H.263相比，无误码环境下，UDVC系统中恢复图像质量下降不超过1.5dB；在5x10E-3误码率环境下，UDVC系统可提高恢复图像质量5dB以上，具有很强的抗误码能力。在本项目资助下，已发表或录用学术论文共13篇(SCI：4篇，EI：9篇)，申请国家技术发明专利11件；有12名研究生参研本项目；参加国际会议或研讨会4次。目前取得的研究成果已经达到项目申请时的预期成果要求，并将在今后1-2年内继续整理输出相关研究成果。

中文关键词： Wyner-Ziv视频编码；边信息；相关噪声模型；率失真模型；鲁棒性重构

英文摘要： Distributed Video Coding can be widely used in wireless communications, thanks to its asymmetric implementation which means the low-complexity at the encoder and its inherent transmission robustness. With the implemented unidirectional distributed video coding (UDVC), the project focused on several aspects including the side information generation algorithm based on the original image correlation, the adaptive correlation noise model generation algorithms, the accurate rate-distortion model construction algorithm of Wyner-Ziv frame DC coefficients, the rate control algorithm at the encoder in UDVC, the robust reconstruction algorithm of the Wyner-Ziv frame at the decoder, the error-resilience method to ensure the quality of Wyner-Ziv frame, distributed video compressed sensing system and so on. By applying above algorithms in UDVC, the experimental results showed that, the encoding time of UDVC is 20% less than that of H.263 codec, and compared with H.263 coding, the average PSNR drop of the restored image by UDVC is less than 1.5dB under error-free environment. Besides, the simulations under the bit-error ratio of 5x10E-3 environment indicated that, UDVC is able to achieve 5dB or higher of the average PSNR gain. Under the support of this project, there are 13 papers (SCI: 4, EI: 9) published or accepted, and 11 patents being applied. 12 students were involved in this project. During the research, we have participated in international conferences or seminars for 4 times. Currently, the project has met the expected goals, and will continue outputting related research results in the next 1-2 years.

英文关键词： Wyner-Ziv video coding; side information; correlation noise model; rate distoration model; robust reconstruction