Improved multi-frame super resolution via multiframe blind deblurring using the alternating direction method of multipliers

摘要

Multi-frame super resolution (MFSR) aims to estimate a high resolution image from a set of low resolution images. Such task is ill-posed and typically computationally costly. We propose a new MFSR forward model, and reformulate the MFSR problem into a problem of multi-frame blind deblurring (MFDB) which is easier to be addressed than the former. We further efficiently solve the MFBD problem via attacking the resulting sub-problems of alternating minimization using the alternating direction method of multipliers (ADMM). Experiments on synthetic and real images show the effectiveness of the proposed method in terms of speed and image quality.