A compact multi-frame blind deconvolution algorithm with the Fourier constraints
Xiangyu Peng, Yonghui Liang · 2021
Multi-frame blind deconvolution (MFBD) is a commonly used method of image post-processing technique to restore high-resolution image from the image observed through ground-based telescopes. During the restoration process, the frequencies with low signal-to-noise (SNR) ratio in the spectrum of the image which are called “spectrum holes” can easily lead to noise amplification effect. Hence there is a filtering idea by calculating the “spectrum holes” to impose the frequency domain constraint in MFBD. Hundreds of images observed through ground-based telescopes are often needed to obtain high quality images, which makes a considerable cost of computation. We discard the information in the images that have not been selected to improve the speed of the process. In this paper, we use the compact blind deconvolution algorithm (CMFBD). A small number of images with better quality as “control frames” is firstly selected to run the traditional MFBD, under such treatment, we can use relative less time to quickly obtain the PSFs corresponding to the “control frames”, and then, the PSFs corresponding to the “non-control frames” is obtained through the “consistency” principle which we assume object is the same in each frame. In CMFBD the non-selected data frames is used to provide an additional constraint on the PSF estimates for the selected data frames.