Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Hierarchical Mixing Architecture for Low-light RAW Image Enhancement

Xianmin Chen, Peiliang Huang, Longfei Han, Dingwen Zhang, Junwei Han

TL;DR

This work tackles the efficiency–quality trade-off in low-light RAW image enhancement by introducing HiMA, a hierarchical architecture that combines Large Scale Blocks in the upper layers and Small Scale Blocks in the lower layers, augmented with Local Distribution Adjustment (LoDA) for adaptive local illumination and Multi-Prior Fusion (MPF) for cross-scale, multi-prior feature fusion. The two-branch design enables robust RAW-to-RGB enhancement while keeping parameter counts low, with three priors produced in the lower branch guiding the upper HiMA branch through MPF. The approach achieves state-of-the-art results on SID, MCR, and ELD datasets with fewer parameters than competing methods, and ablation studies validate the contributions of HiMA, LoDA, and MPF. These findings advance practical RAW-domain enhancement, offering a more efficient solution suitable for resource-constrained devices and real-time applications, with code available for reproducibility.

Abstract

With the rapid development of deep learning, low-light RAW image enhancement (LLRIE) has achieved remarkable progress. However, the challenge that how to simultaneously achieve strong enhancement quality and high efficiency still remains. Leveraging the inherent efficiency of Channel Attention and Mamba, we introduce a Hierarchical Mixing Architecture (HiMA), a hybrid LLRIE framework built upon two core modules. Specifically, we introduce Large Scale Block (LSB) for upper layers and Small Scale Block (SSB) for lower layers that reduce the parameters while improve the performance. Based on this framework, we also introduce a novel Local Distribution Adjustment (LoDA) module that adaptively aligns local feature statistics in a content-aware manner by learning to adjust regional luminance and contrast distributions. Moreover, to alleviate the domain ambiguity commonly observed in existing LLRIE pipelines, we design a Multi-Prior Fusion (MPF) module that leverages three complementary priors extracted from the first stage of the hybrid architecture to maintain domain consistency. Extensive experiments on multiple public benchmarks demonstrate that our approach outperforms state-of-the-art methods, delivering superior performance with fewer parameters. Code is available at https://github.com/Cynicarlos/HiMA.

Hierarchical Mixing Architecture for Low-light RAW Image Enhancement

TL;DR

This work tackles the efficiency–quality trade-off in low-light RAW image enhancement by introducing HiMA, a hierarchical architecture that combines Large Scale Blocks in the upper layers and Small Scale Blocks in the lower layers, augmented with Local Distribution Adjustment (LoDA) for adaptive local illumination and Multi-Prior Fusion (MPF) for cross-scale, multi-prior feature fusion. The two-branch design enables robust RAW-to-RGB enhancement while keeping parameter counts low, with three priors produced in the lower branch guiding the upper HiMA branch through MPF. The approach achieves state-of-the-art results on SID, MCR, and ELD datasets with fewer parameters than competing methods, and ablation studies validate the contributions of HiMA, LoDA, and MPF. These findings advance practical RAW-domain enhancement, offering a more efficient solution suitable for resource-constrained devices and real-time applications, with code available for reproducibility.

Abstract

With the rapid development of deep learning, low-light RAW image enhancement (LLRIE) has achieved remarkable progress. However, the challenge that how to simultaneously achieve strong enhancement quality and high efficiency still remains. Leveraging the inherent efficiency of Channel Attention and Mamba, we introduce a Hierarchical Mixing Architecture (HiMA), a hybrid LLRIE framework built upon two core modules. Specifically, we introduce Large Scale Block (LSB) for upper layers and Small Scale Block (SSB) for lower layers that reduce the parameters while improve the performance. Based on this framework, we also introduce a novel Local Distribution Adjustment (LoDA) module that adaptively aligns local feature statistics in a content-aware manner by learning to adjust regional luminance and contrast distributions. Moreover, to alleviate the domain ambiguity commonly observed in existing LLRIE pipelines, we design a Multi-Prior Fusion (MPF) module that leverages three complementary priors extracted from the first stage of the hybrid architecture to maintain domain consistency. Extensive experiments on multiple public benchmarks demonstrate that our approach outperforms state-of-the-art methods, delivering superior performance with fewer parameters. Code is available at https://github.com/Cynicarlos/HiMA.
Paper Structure (17 sections, 6 equations, 8 figures, 6 tables, 3 algorithms)

This paper contains 17 sections, 6 equations, 8 figures, 6 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Low Light RAW Image Enhancement
  4. Method
  5. Overall Pipeline
  6. Hierarchical Mixing Architecture (HiMA)
  7. Local Distribution Adjustment (LoDA)
  8. Multi-Prior Fusion (MPF)
  9. Experiments
  10. Datasets
  11. Implementation Details
  12. Comparison with SOTA Methods
  13. Ablation Studies
  14. Conclusion
  15. A1. Algorithm details
  16. ...and 2 more sections

Figures (8)

  • Figure 1: Efficiency comparison. Ours method achieves superior performance in balancing number of parameters and PSNR. Even our proposed baseline can also achieve decent performance.
  • Figure 2: Details of the overall pipeline. The original input will pass two branches, the lower one is for pre-denoising and obtaining three priors that will be used in the upper branch. While the upper branch is the main branch with HiMA, and the MPF is used for the skip connection with the priors at the same layer.
  • Figure 3: Details of (a) MeSA and (b) LEB.
  • Figure 4: Motivation of LoDA. The heatmap of the (a) Original noisy RAW, (b) Global Fixed Ratio, (c) Global Aligned Ratio, (d) Local Aligned without Std, (e) Local Aligned with Std, (f) $GT_{RAW}$.
  • Figure 5: The visualization results between our method and the state-of-the-art methods on SID and MCR dataset. (Zoom-in for best view). Note that CANS++ is not compared on MCR dataset since it did not release the pretrained model.
  • ...and 3 more figures