m-Polar fuzzy based image enhancement using exposure fusion

Low light image enhancement is a fundamental pre-processing step in numerous image-processing applications such as medical image analysis, underwater imaging and so on. Many conventional low-light image enhancement methods are typically designed for uniformly dark inputs and tend to over-brighten or distort normally exposed regions in mixed-illumination scene. Since mixed-illumination scenes contain region with differing exposure levels, a single enhancement rule is not adequate; multiple rules must instead be applied jointly within a single image. To address this, we propose a enhancement framework based on m -polar fuzzy set theory where each polar corresponds to a distinct enhancement rule. The input image is transformed into m polar images using parametric fuzzy complement functions, and these polar images are then fused through a multi-scale Laplacian pyramid strategy, with pixel-wise weight maps constructed from contrast, saturation, and well-exposedness. Negative-γ polarities suppress over-brightening in well-lit regions, while positive-γ polarities recover detail in dark areas, enabling selective enhancement without distorting normally exposed content. Experimental evaluation on the TM-DIED and ExDark benchmarks against nine state-of-the-art methods demonstrates that the proposed approach achieves competitive performance across structural similarity index (SSIM), Absolute Mean Brightness Error (AMBE), Lightness Order Error (LOE), Lightness Order Measure (LOM), Structure Measure Operator (SMO), Blind/Referenceless Image Spatial Quality Evaluator (BRISQUE), Entropy, Perception based Image Quality Evaluator (PIQE), Fitness, and Natural Image Quality Evaluator (NIQE) metrics. The source code is available at https://github.com/maheshkumar-cv/m-polar-fuzzy-low-light-image-enhancement .