基于多帧视频序列的盲超分辨率影像重建

Blind Super-resolution Image Reconstruction Based on Multiframe Video Sequence

提出了一种基于多帧序列的盲超分辨率算法和实现框架以解决影像分辨率较低问题。该算法首先对获得的视频序列低分辨率影像进行预处理,然后估计各帧间的相对运动,最后在不知道监控系统的降质模型及相关参数的情况下,逐步投影迭代来重建高分辨率影像。运用该算法对实际监控影像进行了重建实验,通过主观感受和多种客观评价指标多角度对算法进行了评估。结果表明,重建影像空间分辨率得到明显提高,边缘高频信息得到恢复,证明该算法稳健有效。

A blind super-resolution algorithm and the framework based on multiframe are proposed to solve low-resolution problem.Firstly,the observed low-resolution video frames are preprocessed.Then,the relative motions of them are estimated.Finally,the projection iteration is adopted step by step to reconstruct a high-resolution image under the condition but monitoring system model and related parameters are unknown.Some real video monitoring images are used in the experiments and the proposed algorithm is evaluated through subjective feelings and a variety of objective indicators.Results show that the spatial resolutions of reconstructed images are improved and the edge information is recovered.The algorithm is proved to be robust and effective.