现有最好的视频压缩感知重构算法大都采用“预测－ 残差重构”策略，可有效利用帧内和帧间的相关性获得较好的性能，但是残差重构均直接采用SPL 算法，忽略了残
差信号自身的结构特征，限制了性能的进一步提升． 针对该问题，文中提出了一种基于预测残差结构特征的块分类重构算法，首先利用残差块观测值的平均能量对残差块进行分类，然后对不同类的残差块采用不同的重构算法． 仿真实验表明，用于运动较快的视频序列时，文中方案与SPL 算法相比可以获得更好的重构质量．

Most existing compressed video sensing ( CVS) algorithms with best reconstruction performance adopt a “prediction-residual reconstruction”strategy,which helps obtain high reconstruction quality by taking good advantage of intra-frame and inter-frame correlation.However,all of them ignore the residual structure characteristics and simply use SPL reconstruction algorithm which is only suitable for natural image compressed sensing.In order to solve this problem,a block classifying reconstruction algorithm on the basis of residual structure characteristics is proposed,which firstly classifies residual blocks according to their average energy and then adopts suitable algorithms to reconstruct residual blocks corresponding to their structure characteristics.Simulated results show that the proposed algorithm helps achieve higher reconstruction quality than SPL algorithm for video sequences with fast movements.