A Systematic Literature Review of Transmission Estimation Improvements in Dark Channel Prior Digital Image Dehazing

Abstract

Image dehazing, the process of restoring visibility in images degraded by haze, fog, or smoke, is an essential task in computer vision with applications in autonomous driving, remote sensing, and safety-critical environments. Among various methods, the Dark Channel Prior (DCP) has gained widespread attention due to its simplicity and effectiveness in single-image dehazing. However, the original DCP method suffers from limitations such as halo artefacts, sky misestimation, and reduced accuracy in complex scenes. Over the past decade, numerous approaches have been proposed to improve the estimation of the transmission map, which is a critical step in DCP-based dehazing. This systematic literature review (SLR) provides a comprehensive overview of these advancements by analysing 33 selected studies, following PRISMA guidelines. The reviewed methods are categorised into twelve groups, including depth- and scene-adaptive methods, filtering-based refinements, transform-domain approaches, auxiliary image enhancements, physics-based models, statistical and mathematical modelling, feature-based and learning techniques, fusion and multi-scale strategies, region- and boundary-aware methods, colour-channel and prior-based enhancements, optimisation-based methods, and DCP operation improvements. The classification highlights the diversity of strategies used to enhance transmission estimation, ranging from simple filtering and colour-based modifications to complex physics-informed and optimisation-driven approaches. Analysis of the literature indicates that colour-channel and prior-based methods are the most frequently studied, likely due to their simplicity, flexibility, and effectiveness across diverse images. Future research may focus on hybrid approaches that integrate the strengths of multiple strategies, hardware-friendly implementations for real-time applications, and the exploration of richer image features, depth information, or polarisation cues to further improve transmission estimation in challenging environments.