Year: 2022
Type: dissertation
Abstract: In recent years, image defogging has made significant progress. However, many defogging methods either require expensive sensors, or lack the ability to achieve optimal defogging results under all types of foggy conditions using a single model. In this paper, I propose uncloud-bykyleli, an image-based universal image defogging method. uncloud-bykyleli utilizes an unconventional approach that treats foggy areas of an image as irregular partial-holes, and attempts to recover a non-foggy image by using image inpainting techniques on the foggy image. Experiments on benchmark images demonstrate that in comparison to other singular image-based defogging models, uncloud-bykyleli achieves superior performance under all types of foggy-weather conditions.
Author: Li, Kyle
Institution: University of Illinois Urbana-Champaign
Language: English
Topic: Image Enhancement Techniques
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Keywords: image defogging, image inpainting, foggy images
Open Access status: green
OpenAlex ID: W7220791953
