186 | 2 | 32 |
下载次数 | 被引频次 | 阅读次数 |
多帧图像超分辨率重建技术就是将一些变形、模糊、降采样的低分辨率图像进行融合,估计出一幅高分辨率图像,其步骤主要分为运动估计、插值处理、图像重建。本文采用Vandewalle配准方法将处理过的低分辨率图像序列映射到一幅高分辨率网格上,然后进行插值,最后结合小波变换和迭代方法进行图像重建,并采用小波阈值去噪方法进行去噪处理。实验结果表明:本算法能较好地提高图像的峰值信噪比,是图像重建的一种有效方法。
Abstract:Multi-frame super-resolution image reconstruction is used to make image processing produce high resolution image from a set of distortion, blurring and down-sampling images. This technique fuses these low resolution sequences at pixel level to achieve a high resolution image. The steps are as follows: motion estimation, interpolation processing, image reconstruction. The Vandewalle registration method is used to estimate motion parameter, then map a high-resolution grid by interpolation. At last, the algorithm combined with the wavelet and iteration is used to reconstruct the super-resolution image. At the same time, the method of wavelet threshold de-noising is chosen. The experimented results show that this algorithm improves the resolution and PSNR, which is a useful method for super-resolution image reconstruction.
[1]张良培,沈焕锋,张洪艳,等.图像超分辨率重建[M].北京:科学出版社,2012.
[2]孙涛.光学遥感影像复原与超分辨率重建[M].北京:国防工业出版社,2012.
[3]王会鹏.多帧图像超分辨率重建算法研究[D].郑州:解放军信息工程大学,2010.
[4]唐晓初.小波分析及其应用[M].重庆:重庆大学出版社,2005.
[5]PATRICK V,SABINE S,MARTIN V.A frequency domain approach to registration of aliased images with application to super-resolution[J].Applied Signal Processing.2006(4):1-14.
[6]何进荣.基于加权范数迭代算法的总变差正则化图像复原方法[D].武汉:武汉理工大学,2009.
[7]敬忠良,肖刚,李振华.图像融合理论与应用[M].北京:高等教育出版社,2007.
[8]RAYMOND H C T,WAVELET F.Algorithms for high-resolution image reconstruction[J].SIAM J SCI Comput,2003,24:1408-1432.
[9]HONGJIU T,XINJIAN T,JIAN L,et al.Super-resolution remote sensing image processing algorithm based on wavelet transform and interpolation[J].Proceedings of SPIE,2003(48):259-263.
[10]唐磊.多帧图像超分辨率重建算法研究[D].杭州:浙江大学,2011.
[11]张明.图像超分辨率重建和插值算法研究[D].北京:中国科学技术大学,2010.
[12]张兰.多帧图像超分辨率重建关键技术研究[D].天津:天津理工大学,2010.
基本信息:
DOI:
中图分类号:TP391.41
引用信息:
[1]王静,吴爱弟.多帧图像超分辨率重建算法[J].天津职业技术师范大学学报,2014,24(04):32-36.
基金信息:
天津市应用基础及前沿技术研究计划项目(12JCYBJC10600)