Online multi-frame blind deconvolution with super-resolution and saturation correction

Abstract

Astronomical images taken by ground-based telescopes suffer degradation due to atmospheric turbulence. This degradation can be tackled by costly hardware-based approaches such as adaptive optics, or by sophisticated software-based methods such as lucky imaging, speckle imaging, or multi-frame deconvolution. Software-based methods process a sequence of images to reconstruct a deblurred high-quality image. However, existing approaches are limited in one or several aspects: (i) they process all images in batch mode, which for thousands of images is prohibitive; (ii) they do not reconstruct a super-resolved image, even though an image sequence often contains enough information; (iii) they are unable to deal with saturated pixels; and (iv) they are usually non-blind, i.e., they assume the blur kernels to be known. In this paper we present a new method for multi-frame deconvolution called online blind deconvolution (OBD) that overcomes all these limitations simultaneously. Encouraging results on simulated and real astronomical images demonstrate that OBD yields deblurred images of comparable and often better quality than existing approaches.

Multi‐frame blind deconvolution of atmospheric turbulence degraded images with mixed noise models

“…In addition, the observed images are mainly corrupted by Poisson distributed quantum noise and additive Gaussian noise. The Poisson noise is signal-dependent [1], whereas the Gaussian noise is signal-independent [4]. Although Schulz et al proposed to include a mixed noise model earlier [6], their approach is limited because 1) it only works when the parameters of the image system are fully known and 2) actually, their probability model was derived from Poisson noise, different to our approach.…”

“…Many MFBD algorithms and theoretical results have been developed; they used different a priori information in image restoration. Conventional MFBD algorithms usually assume that the observed images are corrupted by a single type of noise, either Poisson noise [1][2][3] or Gaussian noise [4,5]. Instead of adopting these strategies, we propose a novel multi-frame image restoration algorithm by adopting a mixed noise model (MFRAM); MFRAM can achieve a faster convergence, reduce noise more effectively and preserve more image details.…”

Parallel hybrid bispectrum-multi-frame blind deconvolution image reconstruction technique

“…Equation (14) presents multiple paths for the estimation of the object phase at each spatial frequency. To have a reasonable signal to noise ratio (SNR) averaging over five different subplane choices (represented by the value of Df) is sufficient to provide a good balance between execution time and reconstructed image quality for daytime horizontal imaging [10].…”

“…These techniques include least squares minimization [15], simulated annealing [19], genetic algorithm [24], and artificial neural networks [28,29]. In the context of real-time processing, the work in [14] presents an efficient online technique that processes the input sequence one frame at a time. Each new frame is used to gradually improve the image reconstruction.…”

High frame rate imaging based photometry

“…In particular, DIAs utilizing numerical kernels, such as DanDIA (Bramich 2008), seem well-suited because a lucky imaging PSF in general seems to be peculiar and variable with a near diffraction limited core and an extended halo, comparable to the conventional seeing limit (Baldwin et al 2008). In the vein of not wasting photons and cleverly including the high resolution information, another interesting technique is online deconvolution (Hirsch et al 2011). These approaches will be pursued in future work.…”