利用频谱解混叠方法实现超分辨率影像重建

Super-resolution Image Restoration by Spectral De-aliasing

基于信号处理的超分辨率影像重建技术,可以消除由影像系统引起的影像模糊和退化,同时恢复出光学极限外的频谱信息。首先简要介绍了超分辨率影像重建技术的意义和基本过程;并对影像配准与运动模型估计在超分辨率影像重建中的作用给予简要介绍;接着根据连续傅里叶变换(CFT)和离散傅里叶变换(DFT)的频谱混叠关系及CFT位移性质,推导出频谱解混叠的超分辨率影像重建模型;最后采用文献中的运动模型估计方法和本文的重建算法,对几组数据进行实验,获得了空间分辨率提高近1倍的影像。

Superresolution technology is the siginal processing based methods in software. It can remove the blur caused by the imaging system as well as recover spatial frequency information beyond the diffraction limit of the optical system.First this paper gives a brief introduction of superresolution effect and process. And the status and relation of registration and motion estimate in image restoration are presented. Then in order to restore a highresolution image from a sequence of lowresolution, undersampled, discrete frames, this paper presents a new algorithm with reduced the aliasing relationship between the undersampled frames and the reference image. Fullly utilizing the shifts evaluated by the motion estimate that the literature offers, we propose a general formula of restoration in wavenumber domain. Finally success in the use of the algorithm is demonstrated through various simulated examples which afford highresolution image with nearly 2 times resolution as the undersampled frames.