Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Gap-closing Matters: Perceptual Quality Evaluation and Optimization of Low-Light Image Enhancement

Baoliang Chen, Lingyu Zhu, Hanwei Zhu, Wenhan Yang, Linqi Song, Shiqi Wang

TL;DR

This work addresses the lack of perceptual guidance in low-light image enhancement by proposing a gap-closing framework that links subjective quality assessments to model optimization. It introduces SQUARE-LOL, a large-scale paired dataset of enhanced low-light images with pairwise subjective ratings, and a novel BIQA model, IACA, that is illumination-aware and content-adaptive. By replacing traditional priors with IACA-driven perceptual losses in a MAP-like optimization, the approach yields perceptually superior enhancements and strong alignment with human judgments. The framework demonstrates significant improvements in both quality prediction accuracy and the perceptual quality of enhanced images, offering a data-driven pathway to perceptual optimization in low-light imaging.

Abstract

There is a growing consensus in the research community that the optimization of low-light image enhancement approaches should be guided by the visual quality perceived by end users. Despite the substantial efforts invested in the design of low-light enhancement algorithms, there has been comparatively limited focus on assessing subjective and objective quality systematically. To mitigate this gap and provide a clear path towards optimizing low-light image enhancement for better visual quality, we propose a gap-closing framework. In particular, our gap-closing framework starts with the creation of a large-scale dataset for Subjective QUality Assessment of REconstructed LOw-Light Images (SQUARE-LOL). This database serves as the foundation for studying the quality of enhanced images and conducting a comprehensive subjective user study. Subsequently, we propose an objective quality assessment measure that plays a critical role in bridging the gap between visual quality and enhancement. Finally, we demonstrate that our proposed objective quality measure can be incorporated into the process of optimizing the learning of the enhancement model toward perceptual optimality. We validate the effectiveness of our proposed framework through both the accuracy of quality prediction and the perceptual quality of image enhancement. Our database and codes are publicly available at https://github.com/Baoliang93/IACA_For_Lowlight_IQA.

Gap-closing Matters: Perceptual Quality Evaluation and Optimization of Low-Light Image Enhancement

TL;DR

This work addresses the lack of perceptual guidance in low-light image enhancement by proposing a gap-closing framework that links subjective quality assessments to model optimization. It introduces SQUARE-LOL, a large-scale paired dataset of enhanced low-light images with pairwise subjective ratings, and a novel BIQA model, IACA, that is illumination-aware and content-adaptive. By replacing traditional priors with IACA-driven perceptual losses in a MAP-like optimization, the approach yields perceptually superior enhancements and strong alignment with human judgments. The framework demonstrates significant improvements in both quality prediction accuracy and the perceptual quality of enhanced images, offering a data-driven pathway to perceptual optimization in low-light imaging.

Abstract

There is a growing consensus in the research community that the optimization of low-light image enhancement approaches should be guided by the visual quality perceived by end users. Despite the substantial efforts invested in the design of low-light enhancement algorithms, there has been comparatively limited focus on assessing subjective and objective quality systematically. To mitigate this gap and provide a clear path towards optimizing low-light image enhancement for better visual quality, we propose a gap-closing framework. In particular, our gap-closing framework starts with the creation of a large-scale dataset for Subjective QUality Assessment of REconstructed LOw-Light Images (SQUARE-LOL). This database serves as the foundation for studying the quality of enhanced images and conducting a comprehensive subjective user study. Subsequently, we propose an objective quality assessment measure that plays a critical role in bridging the gap between visual quality and enhancement. Finally, we demonstrate that our proposed objective quality measure can be incorporated into the process of optimizing the learning of the enhancement model toward perceptual optimality. We validate the effectiveness of our proposed framework through both the accuracy of quality prediction and the perceptual quality of image enhancement. Our database and codes are publicly available at https://github.com/Baoliang93/IACA_For_Lowlight_IQA.
Paper Structure (21 sections, 3 equations, 9 figures, 4 tables)

This paper contains 21 sections, 3 equations, 9 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Image Quality Assessment Databases
  4. Blind Image Quality Assessment Models
  5. Low-light Image Enhancement Methods
  6. Database Construction
  7. Data Preparation
  8. Subjective Testing
  9. Data Analyses
  10. Methodology
  11. From Enhancement to Assessment
  12. Illumination Modulation
  13. Content Adaptation
  14. From Assessment to Enhancement
  15. Experiments
  16. ...and 6 more sections

Figures (9)

  • Figure 1: Sampled enhanced low-light images in SQUARE-LOL database. The corresponding reference image of each scene is also provided in the first column. The images can be zoomed in to a higher magnification for a closer examination.
  • Figure 2: Illustration of the enhancement results with different methods in the SQUARE-LOL database. The corresponding reference image and low-light image are presented in sub-figure (a) and (l), respectively. The opinion score (OS) is provided below each sub-image.
  • Figure 3: Illustration of the enhancement results with different methods in the SQUARE-LOL database. The corresponding reference image and low-light image are presented in sub-figure (a) and (l), respectively. The opinion score (OS) is provided below each sub-image.
  • Figure 4: The distribution of opinion scores of our SQUARE-LOL database.
  • Figure 5: Illustration of the proposed comprehensive scheme to bridge the gap between quality assessment and enhancement for low-light images. This scheme can be summarized as follows: a) The construction of the valuable SQUARE-LOL database, which provides enhanced low-light images generated using different enhancement (traditional and deep-learning) methods with corresponding quality labels; b) The development of a deep learning-based IQA measure (IACA) dedicated to enhanced images by bridging the gap from enhancement to assessment; c) The optimization of enhancement models with the guidance of the IACA model, filling in the gap from assessment to enhancement.
  • ...and 4 more figures