面向视频传感器网络基于压缩感知理论的多视点视频压缩技术研究

项目名称： 面向视频传感器网络基于压缩感知理论的多视点视频压缩技术研究

项目编号： No.61301174

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 无线电电子学、电信技术

项目作者： 刘翊中

作者单位： 西安电子科技大学

项目金额： 24万元

中文摘要： 融合三维视频的传感器网络能够使人们获得更为丰富的信息内容，在云计算时代有广阔的应用前景。其中对多视点视频图像的压缩编码是这一应用取得成功的关键技术之一。最新理论成果压缩感知具有对信号的低速采样，在采样的同时实现压缩，以及简单的编码器实现架构等特点，适合在能量受限的三维视频传感器网络中对信号进行压缩编码。本项目基于压缩感知理论研究对多视点视频的压缩编码技术。基于冗余字典的压缩感知理论扩展了可稀疏表示的信号范围，基于模型的压缩感知理论利用固定结构的信号模型引导信号恢复，这两种方法更加适用于处理内容更为丰富的信号对象，这符合视频信号的特征。本项目针对多视点视频中的P/B帧，旨在利用帧间相关性引导冗余字典以及图像模型的构建，力求在压缩感知理论的框架下提出多视点视频压缩的创新性编码方法。随着相关研究的不断深入，这种方法将有望在视频传感器网络应用中引起视频图像采样以及压缩编码技术的变革。

中文关键词： 压缩感知；多视点视频；稀疏信号恢复；；

英文摘要： People can acquire a wealth of information by using three-dimensional video based sensor networks, which is rapid with a prosperous application. For this application, it is one of key technologies to compress multiview video effectively. The emerging theory, compressed sensing, supports the features of low speed sampling, compression during sampling, and concise implementation structure for encoder. Therefore, it is proper for compression coding in the energy-restricted three-dimensional video based sensor networks. This project researches on the compressing encode technology of multiview video based on compressed sensing. The overcomplete dictionary based compressed sensing extends the sparse representation ability for extensive signals, and the model-based compressed sensing uses well structured signal models to guide the signal recovery proceeding. Therefore, these two methodologies are more appropriate to processing video signals. This project attempts to use the inter-frame coherence to guide the building of overcomplete dictionary and the image model, and in further to propose innovative encoding technologies for the P/B frames in multiview video based on the theory of compressed sensing. With the intensive study of related topics, this technology is looking forward to introduce great changes to the image

英文关键词： Compressed Sampling；Multiview Video；Sparse Signal Recovery；；