Atmospheric optical turbulence mitigation using iterative image registration and least squares lucky look fusion

Abstract

This paper presents an atmospheric optical turbulence mitigation method that uses a sequence of short-exposure frames. An iterative block matching registration method is proposed for image dewarping. The dewarped frames are combined in a least squares lucky look (LL) fusion process. Here image patches are weighted so as to produce a fused image that is consistent with a theoretical LL optical transfer function (OTF) model. Finally, a Wiener filter is applied to provide deconvolution of the LL OTF. We also explore the LL selectivity tradespace. As the selectivity increases, the LL OTF becomes more favorable but the signal-to-noise ratio suffers, and vice versa. A restoration algorithm is applied to simulated data for quantitative analysis and two different real-world datasets for subjective evaluation. The proposed approach provides improved results compared with the benchmark methods.