Image Deblurring with Algorithm Selection and Evaluation for Real-World Images

EasyChair Preprint 12556, version 1

Abstract

Various techniques for image deblurring have been developed to restore clarity to blurred photographs caused by camera movement. These methods aim to remove blurs and enhance the sharpness of the image, addressing issues such as defocus, motion blur, atmospheric turbulence. In our project, we are tasked with identifying the most effective algorithms for practical applications, including traffic camera images. The evaluation of these algorithms will be based on benchmark scores such as PSNR and SSIM when tested on real blurred photographs. While existing datasets can aid in this evaluation process, it is important to recognize that real-world data may exhibit slight variations. Deblurring techniques can be broadly categorized into blind and non-blind methods, Naf-Net’s, Wiener techniques, deep learning approaches, and hybrid methods. Our primary objective is to apply the best algorithm for deblurring photos to achieve remarkably productive results.

Keyphrases: Gaussian Blur, Wiener model, blind deconvolution, deblurring, hybrid deblurring

@booklet{EasyChair:12556,
  author    = {Hemanth Reddy Somu and Yaswanth Koduri and Jaswanth Prasad Sandhu and Raghava Sai Gunda and Yaswanth Reddy Billa and Vishnu Vardhan Chedella},
  title     = {Image Deblurring with Algorithm Selection and Evaluation for Real-World Images},
  howpublished = {EasyChair Preprint 12556},
  year      = {EasyChair, 2024}}