传统的图像复原一般认为点扩散函数(PSF)是空间不变的, 实际光学系统由于受到像差等因素的影响, 并非严格的线性空间不变系统, 基于空间变化PSF的非盲去卷积图像复原法逐渐体现其优越性。空间变化PSF的非盲去卷积图像复原法先准确估计图像空间变化的PSF, 再利用非盲去卷积算法对图像进行复原, 有利于恢复出高质量图像。本文从算法的角度综述了近几年提出的基于空间变化PSF的非盲去卷积图像复原方法, 并对比了基于强边缘预测估计PSF的非盲去卷积法、基于模糊噪声图像对PSF估计非盲去卷积法等算法的优缺点, 各算法分别在PSF估计精确度、振铃效应抑制效果、适用范围等方面体现出各自的优劣。空间变化PSF的非盲去卷积图像复原法的研究, 有利于推进图像复原技术向更高水平发展, 使光学系统往轻小型化方向发展, 从而在多个科学领域发挥其重要作用。

Traditional image restoration is generally considered that point spread function(PSF) is space-invariant. However, the actual optical system suffering from various optical aberrations can not be strictly linear space invariant. Non-blind deconvolution(NBD) algorithm of image restoration based on spatially-varying PSF(SVPSF) gradually embodies its superiority. NBD image restoration with SVPSF accurately estimates the spatially-varying PSF of the image at first, and then restores the image through NBD algorithm, which is conducive to the recovery of high quality images. From the perspective of algorithm, we review non-blind image restoration method proposed in recent years based on spatially-varying PSF, as well as compare merits and drawbacks among NBD algorithm based on PSF estimation using sharp edge prediction, NBD algorithm based on blurred/noisy image pairs, and so on. These algorithms reflect pros and cons respectively in PSF estimation accuracy, inhibitory effect of ringing artifacts, and the scope of application. The study of the NBD image restoration method based on SVPSF is beneficial to the development of image restoration technology to a higher level, which facilitates the optical systems to be smaller, so that it can play an important role in many scientific fields.