多媒体传感器网络视频的帧分类联合重构

Frame Classification Based Joint Video Recovery Algorithm for Multimedia Sensor Networks

为降低多媒体传感器网络中视频压缩感知的计算复杂度,提出一种基于帧分类的多媒体传感器网络视频联合重构算法。依据视频数据的联合稀疏模型将视频帧分为关键帧和非关键帧。对于压缩感知重构中欠定线性方程组,可利用关键帧和非关键帧之间的相关边信息进行重构初始化,同时运用有界约束二次规划对其进行求解。从仿真结果可知,相对于传统的视频压缩感知算法而言,在保证视频重构质量的前提下,所提方法在重构算法复杂度上不但能有效降低,同时,在视频重构上能提高其实时性。

For reducing the computation complexity of compressed sensing （CS） based video recovery algorithm in multimedia sensor networks, a frame classification based joint recovery algorithm is proposed. The frames of network video are classified into two types, which are key frames and non-key frames, and the side information between two types of frames can be used in initialization of recovery. The under-determined linear equations in CS recovery are solved by bound-constrained quadratic programming. Simulation results demonstrate that the proposed algorithm can significantly reduce the complexity while the accuracy of recovery is ensured, thus im- prove the real-time performance of video recovery in multimedia sensor networks.