FLARE-BO: Fused Luminance and Adaptive Retinex Enhancement via Bayesian Optimisation for Low-Light Robotic Vision

TL;DR

FLARE-BO is a Bayesian optimisation-based framework that enhances low-light robotic vision by jointly tuning eight image processing parameters, significantly improving image quality without training.

Contribution

The paper introduces FLARE-BO, an extended Bayesian optimisation framework that optimally adjusts eight parameters for low-light image enhancement, surpassing previous limited approaches.

Findings

Marked improvements over existing methods on the LOL dataset.

Joint optimisation of multiple parameters enhances image quality.

No training required, enabling adaptable real-time enhancement.

Abstract

Reliable visual perception under low illumination remains a core challenge for autonomous robotic systems, where degraded image quality directly compromises navigation, inspection, and various operations. A recent training free approach showed that Bayesian optimisation with Gaussian Processes can adaptively select brightness, contrast, and denoising parameters on a per-image basis, achieving competitive enhancement without any learned model. However, that framework is limited to three parameters, applies no illumination decomposition or white balance correction, and relies on Non-Local Means denoising, which tends to over smooth edges under noisy conditions. This paper proposes FLARE-BO (Fused Luminance and Adaptive Retinex Enhancement via Bayesian Optimisation), an extended framework that jointly optimises eight parameters spanning across gamma correction, LIME-style illumination normalisation, chrominance denoising, bilateral filtering, NLM denoising, Grey-World automatic white balance, and adaptive post smoothing. The search engine employs a unit hypercube parameter normalisation, objective standardisation, Sobol quasi-random initialisation, and Log Expected Improvement acquisition for principled exploration of the expanded space. Performance of the proposed method is benchmarked using the Low Light paired dataset (LOL) and results show marked improvements of the proposed method over existing methods that were not specifically trained using this dataset.