A comparative study of low-light fingerprint image enhancement methods based on multimetric evaluation
Abstract
Low-light fingerprint images often suffer from insufficient illumination, low ridge-valley contrast, and missing details, which severely compromise the accuracy of minutiae extraction in Automated Fingerprint Identification Systems (AFIS) and threaten the reliability of forensic identification. While deep learning-based low-light image enhancement methods have achieved remarkable progress in natural image processing, their "black-box" nature and reliance on large-scale paired training data limit their direct application in forensic scenarios where hallucination risks and data scarcity are critical concerns. To address this gap, this study constructs a standardized multi-metric evaluation framework for low-light fingerprint enhancement, systematically comparing six representative traditional algorithms: CLAHE, Multiscale Retinex (MSR), Gamma correction, global histogram equalization (HistEq), homomorphic filtering, and bilateral filtering. The evaluation integrates nine metrics across three dimensions: fidelity (PSNR, SSIM, MSE, Brightness Error), clarity (Average Gradient, Contrast, Entropy), and naturalness (PSNR-B, NIQE). Experimental results demonstrate that the MSR algorithm achieves optimal overall performance, delivering the highest PSNR (8.31 dB), SSIM (0.49), and Entropy (7.12), while maintaining superior naturalness (NIQE: 1.47). Although CLAHE exhibits advantages in enhancing local contrast, its over-amplification of background textures may interfere with AFIS matching. In contrast, Gamma, homomorphic filtering, and bilateral filtering fail to effectively restore fingerprint details under default parameters. Furthermore, through extended parameter sensitivity analysis, this study identifies stable operating intervals for MSR scale parameters and CLAHE clip limits, providing actionable guidance for forensic practitioners. This study provides empirical evidence for selecting robust enhancement methods in forensic practice, balancing detail preservation and visual naturalness for evidentiary integrity.