Improving Neural Blind Deconvolution

0546240 - ÚTIA 2022 RIV US eng C - Konferenční příspěvek (zahraniční konf.)
Kotera, Jan - Šmídl, Václav - Šroubek, Filip
Improving Neural Blind Deconvolution.
2021 IEEE International Conference on Image Processing : Proceedings. Piscataway: IEEE, 2021, s. 1954-1958. ISBN 978-1-6654-4115-5. ISSN 2381-8549.
[IEEE International Conference on Image Processing (ICIP) 2021. Anchorage (US), 19.09.2021-22.09.2021]
Grant CEP: GA ČR GA20-27939S
Institucionální podpora: RVO:67985556
Klíčová slova: blind deblurring * SelfDeblur * deep image prior
Obor OECD: Robotics and automatic control
http://library.utia.cas.cz/separaty/2021/ZOI/kotera-0546240.pdf

The field of blind image deblurring was for a long time dominated by Maximum-A-Posteriori methods seeking the optimal pair of sharp image--blur of a suitable functional. Recently, learning-based methods, especially those based on deep convolutional neural networks, are proving effective and are receiving increasing attention by the research community. In 2020, Ren~et~al. proposed a deblurring method called SelfDeblur which combines the model-driven approach of traditional MAP methods and the generative power of neural nets. The method is capable of producing very high-quality results, yet it inherits some problems of MAP methods, especially possible convergence to a wrong local optimum. In this paper we propose several easy-to-implement modifications of SelfDeblur, namely suitable initialization, multiscale processing, and regularization, that improve the average performance of the original method and decrease the probability of failure.
Trvalý link: http://hdl.handle.net/11104/0322888