Model-Based Single Image Deep Dehazing

TL;DR

This paper introduces a hybrid dehazing method that combines model-based initialization with deep learning refinement to effectively remove haze while preserving image details.

Contribution

It proposes a novel fusion of model-based and data-driven techniques for single image dehazing, improving haze removal and image quality.

Findings

Effective haze removal on real-world images

High-quality dehazed images with preserved details

Outperforms traditional methods in visual quality

Abstract

Model-based single image dehazing algorithms restore images with sharp edges and rich details at the expense of low PSNR values. Data-driven ones restore images with high PSNR values but with low contrast, and even some remaining haze. In this paper, a novel single image dehazing algorithm is introduced by fusing model-based and data-driven approaches. Both transmission map and atmospheric light are initialized by the model-based methods, and refined by deep learning approaches which form a neural augmentation. Haze-free images are restored by using the transmission map and atmospheric light. Experimental results indicate that the proposed algorithm can remove haze well from real-world and synthetic hazy images.