Image Super Resolution for Vehicle Number Plate Enhancement

  • Prof.(Dr) Rajni Jindal, Sparsh Shubham, Shubham Yadava, Anshika Singh

Abstract

In this paper, vehicle licence plate clarity using super resolution and DeblurGAN is implemented. As an alternative to using expensive hardware for increasing vehicle number plate readability, we propose the implementation of a super resolution approach to upscale low-resolution images using SR-GANs. The model also uses a plate detection and warping algorithm for ensuring less noise goes into the model. We also implement DeblurGANs to remove blur from the upscaled image. The model will finally generate a character string of what its best prediction is using an OCR model. This approach is implemented instead of simply using an OCR because at low resolutions, characters such as {2, Z}, {0, O, Q}, {1,7}, etc. are difficult to distinguish between. For multiple frames, an election algorithm will give the most probable output based on outputs of each frame.

Published
2020-07-01
How to Cite
Prof.(Dr) Rajni Jindal, Sparsh Shubham, Shubham Yadava, Anshika Singh. (2020). Image Super Resolution for Vehicle Number Plate Enhancement. International Journal of Advanced Science and Technology, 29(7), 13526 - 13546. Retrieved from http://sersc.org/journals/index.php/IJAST/article/view/29461
Section
Articles