基于自适应滤波的视频序列超分辨率重建

针对视频序列的超分辨率重建,提出了一种动态自适应滤波方法.在最大后验概率估计和加权最小二乘的基础上,给出视频序列超分辨率重建数学模型;深入研究了运动补偿矩阵和权值矩阵的构成和性质;详细推导了自适应滤波器的递推公式;分析了算法的存储与计算复杂度.仿真实验表明该算法的重建结果相当有效,相比双三次插值和无运动补偿的单帧迭代重建,可以获得一定的PSNR增益;与Elad滤波方法相比,具有更小的计算量和更强的自适应性和鲁棒性.

Super Resolution Reconstruction of Video Sequence Based on Adaptive Filter

A dynamic adaptive filter for video sequence super resolution reconstruction is proposed. Firstly, the super resolution reconstruction model for video sequence is put forward based on maximum a-posteriori probability and weighted least square. Then, the authors thoroughly study the motion compensation matrix and weighted matrix and deduce the recursive filter formula. Finally, the algorithm complexity has also been analyzed in detail. Experimental results demonstrate the power of the proposed method. On the one hand, some improvement in PSNR is attained by using the adaptive filter in place of bicubic interpolation and single frame reconstruction without motion compensation. On the other hand, compared with the Elad method, the presented method has smaller computation and stronger adaptability and robustness.