基于L曲线的自适应时空正则化视频超分辨率重建

Video Super-Resolution Reconstruction Based on L-curve Adaptive Spatial-Temporal Regularization

为了解决视频超分辨率重建的病态问题,以得到良好的重建效果,提出了一种新颖的视频超分辨率重建算法。在算法中引入了时空联合正则化算子,通过视频帧本身的空间平滑信息和视频相邻帧的帧间相关先验信息的引入,提高了解的质量;同时,为了选择合适的时空正则化系数,提出了基于L曲线的自适应时空正则化系数计算方法,可以自适应地计算合适的正则化系数。通过对模拟图像序列和真实视频序列的实验结果表明,算法能得到较为精确的解,重建出具有良好视觉效果的高分辨率视频。

In order to solve the ill-posed problem of super-resolution reconstruction and achieve good visu- al effect, this paper proposes a novel video super-resolution reconstruction algorithm. In this algorithm, spatial-temporal regularization operator is introduced in the reconstruction process, and the precision of solving is improved by spatial smoothing information and prior information between video frames. In addition, a method of adaptively computing spatial-temporal regularization factor is proposed in this algorithm, and it is based on L-curve and can adaptively compute an appropriate regularization factor. Experimental results of simulated video sequence and real video sequence prove that the proposed super-resolution reconstruction algorithm can obtain precise solving and high-resolution visual result simultaneously.