An intelligent low-light image enhancement model using intuitionistic fuzzy generator and genetic algorithm validated through MCDM

Low-light image enhancement (LLIE) plays a crucial role in various computer vision and pattern recognition applications, including surveillance, biometric authentication, and security systems. Images captured under poor illumination often suffer from low contrast, shadows and uneven brightness, resulting in loss of critical information. To address these challenges, this study proposes a novel LLIE model based on a new intuitionistic fuzzy generator (IFG) and genetic algorithm (GA) optimization. The model begins by fuzzifying the input image using a linear membership function to capture the degree of intensity for each pixel. Based on this, a new IFG is derived to compute both the membership and non-membership values, thereby generating an intuitionistic fuzzy image that models uncertainty more effectively. To further improve the contrast and brightness of the image, the contrast-limited adaptive histogram equalization (CLAHE) method is applied after the LAB and HSV color space conversion. The outputs are fused using principal component analysis (PCA) and GA is employed to optimize the fused image based on entropy maximization. The final image is obtained through defuzzification, the reverse process of the initial fuzzification process. Quantitative analysis on the LOL dataset shows that the proposed model achieves SSIM of 0.5842, PSNR of 18.29 dB and entropy of 7.5531, outperforming several existing and state-of-the-art deep learning-based methods. A multi-criteria decision-making (MCDM) method, TOPSIS, is applied to integrate eight conflicting performance metrics and rank 14 enhancement techniques, further confirming the superiority of the proposed model. Furthermore, the ablation study shows the efficiency of the proposed fusion-based model.