FLARE-BO: Fused Luminance and Adaptive Retinex Enhancement via Bayesian Optimisation for Low-Light Robotic Vision
FLARE-BO is a training-free low-light image enhancement framework that uses Bayesian optimization to jointly tune eight parameters for complex image processing, including LIME illumination normalization and bilateral filtering. It outperforms existing methods on the Low Light paired dataset by achieving significant improvements in structural similarity and reconstruction quality, making it ideal for autonomous robotic vision in uncontrolled lighting.