Enhancing Visibility: Multiresolution Dark Channel Prior for Dehazing and Fog Removal in Images

Abstract

Capturing images in conditions marked by fog or smog results in compromised visual quality, characterized by reduced visibility and contrast. Such limitations impede critical tasks like image segmentation, target detection, and video surveillance within outdoor monitoring systems. This paper presents an effective image defogging algorithm designed to rectify these issues and restore image clarity. In this study, we introduce a method centered around a dark channel prior, a foundational image characteristic preceding the haze removal process. This prior harnesses statistical insights from haze-free outdoor images, revealing a significant finding: numerous set of patches in haze-free images harbor pixels exhibiting remarkably minuscule intensities in more than a single colour channel. Through integration of this algorithm into the imaging model for hazing, the algorithm adeptly estimates haze thickness, thereby facilitating the recovery of high-quality, haze-free images. We take a novel approach enhancing the dark channel prior, and its practical implementation demonstrates promise. By improving visibility and contrast, it has the potential to enhance the performance of outdoor monitoring systems, including video surveillance, in unfavorable weather conditions. Extensive testing underscores the effectiveness of our approach in improving image quality and its utility across various real-world applications.

Key Findings

• Enhanced the Dark Channel Prior (DCP) algorithm to more accurately estimate haze thickness in outdoor images.
• Demonstrated significant improvements in visibility and contrast for video surveillance systems in poor weather.
• Validated that haze-free patches consistently exhibit low intensity in at least one color channel, enabling robust haze removal.
• Proposed a computationally efficient method suitable for real-time applications like autonomous driving and security monitoring.