压缩视频超分辨率重建技术.pdfVIP

第 29 卷第2 期 电子与信息学报 Vo 1.29No.2 2007 年2 月 Journal of Electronics &Information Technology Feb. 2007 压缩视频超分辨率重建技术 徐忠强 朱秀昌 (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) 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，而且三缩视顿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、低分辨来图像间的关系，给:H'r 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、 rI~ 集投影(POCS)算法和法代反向投影(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。 关键词:压缩视频:超分辨率:最大后验概率; r1l 集投影:选代反IÎIJ 投影 中图分类号: TP391 文献标识码:A 文章编号: 1009-5896(2007)02-0499-07 Super-resolution Reconstruction Technology for Compressed Video Xu Zhong-qiang Zhu Xiu-chang (Information Indust -ry Ministry and Jiangsu Province Key Lab of lrrωge Processing & Image Communication, Nanjing Unive'T'sity of Posts &l'elecornmunications, Nαη， :jing 210003 , China) Abstract: Supcr-Rcsolutioll (SR) rcconstruction is to cstimate High-Rcsolution (HR) imagcs from Low-Rcsolution (LR) image scquencc , which has becn a grcat focus for compre8scd vidco. This paper firstly prescnts the base of SR rcconstruction for compresscd vidco by building the relatiollships bctwcen thc HR and LR images and survcying the models of qllantization noise and motion vector. Then ,thc typical algorithms 8uchωMAP， POCS and IBP arc described in detail with experimental results. The computation complexity and real-tirnc approachcs are also investigated. Finally, aiming at the drawback8 , thc rescarch rang of prospects are dernonstrated through pointing out thc focus on dcgradation mode , rnotion estimation , rcconstruction algorithm and real-timc irnplcmcntations. Key words: Cornp