Title: Single-image reflection removal algorithms: a systematic review using PRISMA guidelines

Authors: Hui Hu; Wai Chong Chia; Yunli Lee; Kok-Lim Alvin Yau; Han Huang

Addresses: Faculty of Engineering and Technology, Sunway University, Bandar Sunway, 47500, Malaysia ' Faculty of Engineering and Technology, Sunway University, Bandar Sunway, 47500, Malaysia ' Faculty of Engineering and Technology, Sunway University, Bandar Sunway, 47500, Malaysia ' Lee Kong Chian Faculty of Engineering and Science, Universiti Tunku Abdul Rahman (UTAR), Selangor, 43000, Malaysia ' School of Software Engineering, South China University of Technology, Guangzhou, 510006, China

Abstract: Taking photos through glasses or windows often introduces reflections that affect the accuracy of computer vision tasks. This systematic literature review provides a comprehensive survey on single-image reflection removal for general scenes captured through glasses. We present the priors, the factors involved in the mixture image formation process, and the quantitative metrics in model-driven methods, as well as the training and testing datasets and quantitative metrics in data-driven methods. Our review addresses a total of three research questions, ranging from 2017 to 2023, in accordance with the PRISMA 2020 guidelines. We screened over 566 research papers from four electronic databases - ScienceDirect, IEEE Xplore, Web of Science, and the ACM Digital Library - and ultimately selected 36 papers for in-depth analysis. By comprehensive analysis and statistics of the selected papers, we answered the 3 key research questions and then identified open problems for future researchers.

Keywords: reflection removal; reflection separation; deep learning; single-image reflection removal; systematic literature review.

DOI: 10.1504/IJCSM.2025.148204

International Journal of Computing Science and Mathematics, 2025 Vol.21 No.4, pp.355 - 367

Received: 14 Aug 2024
Accepted: 29 May 2025
Published online: 29 Aug 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article