Entropy-Driven Genetic Optimization for Deep-Feature-Guided Low-Light Image Enhancement

TL;DR

The paper tackles low-light image enhancement with a focus on semantic fidelity by proposing an unsupervised, fuzzy-inspired framework that optimizes three image parameters (brightness $b$, contrast $c$, and gamma $\gamma$) using a GPU-accelerated NSGA-II. It couples a frozen deep feature extractor $\mathcal{F}$ with entropy maximization and deep-feature alignment to balance perceptual quality and semantic content, and augments the search with a memetic local search. The approach operates without paired training data and demonstrates competitive to state-of-the-art performance on unpaired datasets (via BRISQUE/NIQE) and strong results on paired MIT-5K (via PSNR/SSIM/Entropy), while offering transparency and tunable constraints. The work advances unsupervised LLIE by integrating multi-objective optimization, deep-feature guidance, and per-image parameterization, with practical implications for domains lacking labeled data.

Abstract

Image enhancement methods often prioritize pixel level information, overlooking the semantic features. We propose a novel, unsupervised, fuzzy-inspired image enhancement framework guided by NSGA-II algorithm that optimizes image brightness, contrast, and gamma parameters to achieve a balance between visual quality and semantic fidelity. Central to our proposed method is the use of a pre trained deep neural network as a feature extractor. To find the best enhancement settings, we use a GPU-accelerated NSGA-II algorithm that balances multiple objectives, namely, increasing image entropy, improving perceptual similarity, and maintaining appropriate brightness. We further improve the results by applying a local search phase to fine-tune the top candidates from the genetic algorithm. Our approach operates entirely without paired training data making it broadly applicable across domains with limited or noisy labels. Quantitatively, our model achieves excellent performance with average BRISQUE and NIQE scores of 19.82 and 3.652, respectively, in all unpaired datasets. Qualitatively, enhanced images by our model exhibit significantly improved visibility in shadowed regions, natural balance of contrast and also preserve the richer fine detail without introducing noticable artifacts. This work opens new directions for unsupervised image enhancement where semantic consistency is critical.

Figures (8)

• Figure 1: Top-level diagram of the framework
• Figure 2: Visual comparison of enhancement results by our method versus CIDNet on a random sample from each dataset
• Figure 3: Visual comparison of all the approaches-(a)-retouched version ,(b)-CIDNet lolv1 weight , (c)-CIDNet generalization weight, (d)-ours, (e)- actual image
• Figure 4: Example enhancement on the DICM dataset using our proposed method
• Figure 5: Example enhancement on the LIME dataset using our proposed method