Blind deconvolution with automatic relevance determination model based on the kernel gradient

Yu Ma; Xin-He Wang; Jun-Jun Xiao

doi:10.1117/12.2577161

7 October 2020 Blind deconvolution with automatic relevance determination model based on the kernel gradient

Yu Ma, Xin-He Wang, Jun-Jun Xiao

Proceedings Volume 11571, Optics Frontier Online 2020: Optics Imaging and Display; 1157108 (2020) https://doi.org/10.1117/12.2577161
Event: Optics Frontiers Online 2020: Optics Imaging and Display (OFO-1), 2020, Shanghai, China

Abstract

In this work, based on the principle of blind deconvolution, and considering the inherent structure of the blur kernel, an automatic relevance determination (ARD) model is used to determine the prior model based on the gradient, instead of intensities of the blur kernel. The results show that the proposed ARD model on gradient can improve the kernel quality and thus produce better de-blurred image. Compared with the case of using the intensity for ARD, the proposed algorithm gives better blur kernel estimation and show robustness against non-Gaussian noise. As concrete examples, we demonstrate that the proposed method is applicable to image restoration on the scenario of camera shake, object motion and defocused blurring.

Citation Download Citation

Yu Ma, Xin-He Wang, and Jun-Jun Xiao "Blind deconvolution with automatic relevance determination model based on the kernel gradient", Proc. SPIE 11571, Optics Frontier Online 2020: Optics Imaging and Display, 1157108 (7 October 2020); https://doi.org/10.1117/12.2577161

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available