Efficient Hybrid Evolutionary–Numerical Algorithms for Contrast Enhancement Under Distortion Constraints in Medical Imaging

Image contrast enhancement is widely used to improve visual perception in digital images; however, it often amplifies noise and introduces artifacts that distort structural information. To address this issue, CLAHE-based contrast enhancement is formulated as a constrained optimization problem, in which distortion control is enforced via PSNR constraints. In this work, a behavioral analysis of the decision variables is conducted, revealing distinct objective-function responses that are exploited to guide the optimization process. Based on these observations, a hybrid evolutionary–numerical framework is developed, combining evolutionary search for discrete parameter exploration with numerical optimization for stable adjustment of continuous parameters. The proposed methods are evaluated on a benchmark set of 30 medical images and compared against fully evolutionary, numerical, and recent population-based optimization approaches reported in the literature. Experimental results show that the hybrid variants, particularly NR-EVO, consistently achieve the best overall performance across different computational budgets, producing higher-quality enhancements for the evaluated benchmark problems. On average, the enhanced images exhibit an increase in entropy of approximately 22% while maintaining competitive structural similarity and satisfying the predefined distortion constraints.