基于混合POCS的视频图像超分辨率重建

提出混合凸集投影算法HPOCS对视频图像进行超分辨率重建,利用连续视频图像间的互异信息生成更高分辨率的视频图像.该算法首先采用图像匹配估计视频图像间的运动位移;然后进行基于的APEX盲解卷积,估计点扩散函数和理想视频图像;最后在凸集投影的理论框架下进行图像重建.实验表明,HPOCS重建后,视频图像的分辨率相对于原始图像、双线性内插图像和POCS重建图像明显提高,图像边缘更加清晰,细节信息更加突出.

Superresolution Video Reconstruction with a Hybrid POCS Algorithm

In the paper the Hybrid Projection onto to Convex Sets (HPOCS) scheme for video superresolution reconstruction is proposed, which aims to integrate slightly different, but unique, information in temporally adjacent video frames to generate much higher resolution reconstruction frames. In the proposed algorithm, image matching is utilized to estimate translational motion among various frames firstly, and then APRX algorithm is adopted to access the Point Spread Function (PSF) and the ideal high resolution (HR) video frame, finally, image reconstruction is carried out within the projection onto convex sets framework. Experimental results computed from simulation video sequence and actual video sequences show dramatic resolution improvement over original frame, bilinear interpolation frame and POCS reconstruction result. The edges in the HPOCS reconstruction result are much clearer and the details are more prominent