Multi-Branch Non-Homogeneous Image Dehazing via Concentration Partitioning and Image Fusion

2026-04-27

AI Summary

This work addresses the challenge that existing single-image dehazing methods struggle to effectively handle non-uniform haze characterized by spatially varying concentrations and abrupt density transitions. To this end, the authors propose CPIFNet, a novel framework that, for the first time, models a non-uniform hazy image as multiple approximately uniform local regions. The approach employs a multi-branch enhancement network (IENet) to separately process regions with differing haze densities and subsequently integrates the outputs through an image fusion network (IFNet) that adaptively combines the strengths of each branch. By incorporating deep feature stacking and a composite loss function—comprising reconstruction, perceptual, structural, and color consistency terms—the method significantly enhances dehazing quality, achieving superior performance in detail preservation and color fidelity.

Abstract

Existing single image dehazing methods have demonstrated satisfactory performance on homogeneous thin-haze images; however, they often struggle with non-homogeneous hazy images that exhibit spatially varying haze concentrations and abrupt density transitions across different regions. To address this fundamental limitation, we propose a novel multi-branch deep neural network framework, termed Concentration Partitioning and Image Fusion Network (CPIFNet), which decomposes the challenging non-homogeneous dehazing problem into a set of tractable homogeneous sub-problems. Our key insight is that a single non-homogeneous hazy image can be viewed as a composite of multiple local regions, each exhibiting approximately homogeneous haze characteristics. CPIFNet employs a two-stage architecture consisting of an Image Enhancement Network (IENet) stage and an Image Fusion Network (IFNet) stage. In the first stage, multiple IENet branches are independently trained on homogeneous haze datasets of different concentration levels, producing enhancement models that excel at restoring regions matching their respective haze densities. In the second stage, the IFNet intelligently aggregates the advantageous regions from all enhancement outputs through deep feature stacking and merging, yielding a unified high-quality dehazed result. Furthermore, we introduce a comprehensive loss function incorporating reconstruction, perceptual, structural, and color losses to jointly supervise both stages.

Problem

• non-homogeneous dehazing
• spatially varying haze
• density transitions
• image dehazing
• haze concentration

Innovation

• non-homogeneous dehazing
• concentration partitioning
• multi-branch network
• image fusion
• deep neural network