Publication detail

CNN for license plate motion deblurring

SVOBODA, P. HRADIŠ, M. MARŠÍK, L. ZEMČÍK, P.

Original Title

CNN for license plate motion deblurring

Type

conference paper

Language

English

Original Abstract

In this work we explore the previously proposed approach of direct blind deconvolution and denoising with convolutional neural networks (CNN) in a situation where the blur kernels are partially constrained. We focus on blurred images from a real-life traffic surveillance system, on which we, for the first time, demonstrate that neural networks trained on artificial data provide superior reconstruction quality on real images compared to traditional blind deconvolution methods. The training data is easy to obtain by blurring sharp photos from a target system with a very rough approximation of the expected blur kernels, thereby allowing custom CNNs to be trained for a specific application (image content and blur range). Additionally, we evaluate the behavior and limits of the CNNs with respect to blur direction range and length.

Keywords

Convolutional neural network, Motion blur, Image reconstruction, Blind deconvolution, License plate

Authors

SVOBODA, P.; HRADIŠ, M.; MARŠÍK, L.; ZEMČÍK, P.

Released

25. 9. 2016

Publisher

IEEE Signal Processing Society

Location

Phoenix

ISBN

978-1-4673-9961-6

Book

IEEE International Conference on Image Processing (ICIP)

Pages from

3832

Pages to

3836

Pages count

4

URL

http://ieeexplore.ieee.org/document/7533077/

BibTex

@inproceedings{BUT133500,
  author="Pavel {Svoboda} and Michal {Hradiš} and Lukáš {Maršík} and Pavel {Zemčík}",
  title="CNN for license plate motion deblurring",
  booktitle="IEEE International Conference on Image Processing (ICIP)",
  year="2016",
  pages="3832--3836",
  publisher="IEEE Signal Processing Society",
  address="Phoenix",
  doi="10.1109/ICIP.2016.7533077",
  isbn="978-1-4673-9961-6",
  url="http://ieeexplore.ieee.org/document/7533077/"
}