压缩视频超分辨率重建技术  

Super-resolution Reconstruction Technology for Compressed Video

作　　者：徐忠强[1] 朱秀昌[1] 

机构地区：[1]南京邮电大学"图像处理与图像通信"信息产业部和江苏省重点实验室,南京210003

出　　处：《电子与信息学报》2007年第2期499-505,共7页Journal of Electronics & Information Technology

基　　金：江苏省自然科学基金(BK2004151)资助课题

摘　　要：超分辨率图像重建就是由低分辨率图像序列来估计高分辨率图像,而压缩视频的重建正成为当前研究的热点。本文首先分析了压缩视频重建的基础,建立了高、低分辨率图像间的关系,给出了量化噪声和运动矢量的模型;接着对目前最具有代表性的最大后验概率(MAP)算法、凸集投影(POCS)算法和迭代反向投影(IBP)算法进行了详细的阐述,并分别给出了实验结果;然后分析了运算的复杂度,介绍了几种实时化方法;最后针对目前存在的问题进行了展望,指出降质模型、运动估计、重建算法和实时应用将是今后研究的重点。Super-Resolution （SR） reconstruction is to estimate High-Resolution （HR） images from Low-Resolution （LR） image sequence, which has been a great focus for compressed video. This paper firstly presents the base of SR reconstruction for compressed video by building the relationships between the HR and LR images and surveying the models of quantization noise and motion vector. Then, the typical algorithms such as MAP, POCS and IBP are described in detail with experimental results. The computation complexity and real-time approaches are also investigated. Finally, aiming at the drawbacks, the research rang of prospects are demonstrated through pointing out the focus on degradation mode, motion estimation, reconstruction algorithm and real-time implementations.

关 键 词：压缩视频 超分辨率 最大后验概率 凸集投影 迭代反向投影 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]