Channel-Selective Retinal Image Enhancement in Lab Color Space UsingICF-Guided Rank-Preserving CLAHE

Enhancement of retinal fundus images is essential for improving the visibility of subtle diagnostic features that are often obscured by poor contrast, uneven illumination, and background artifacts. To address these limitations, this paper presents a contrast-aware enhancement framework based on an Integrated Contrast Feature (ICF) metric, developed to selectively improve diagnostically significant structures in the Lab color space while preserving naturally well-contrasted regions.The proposed framework begins with Region-of-Interest (ROI) extraction to remove irrelevant background areas and focus solely on the retinal field. For each channel (L, a, b), the ICF is computed by combining localized spatial contrast with wavelet-derived frequency energy, producing a reliable indicator of contrast quality. Each channel is then evaluated against statistically derived thresholds obtained from Gaussian modeling of annotated datasets, classifying it as either low-contrast or sufficient-contrast.Low-contrast channels are selectively enhanced using two proposed methods: Contrast-Aware Selective CLAHE (CAS-CLAHE) and Contrast-Aware Selective Rank-Preserving CLAHE (CAS-RP-CLAHE). The latter scheme further preserves the local rank order of intensities, preventing luminance reversal and ensuring continuity of delicate retinal structures such as microvessels.Extensive experiments on benchmark datasets demonstrate that CAS-RP-CLAHE achieves the best overall performance, delivering perceptual quality scores of NIQE = 3.63, CEIQ = 3.74, and CII = 1.90. Independent evaluation by ophthalmologists confirmed that the method enhances critical diagnostic regions—including the optic disc, vascular network, and exudates—with improved clarity and without introducing unnatural artifacts.By strengthening visual quality while maintaining diagnostic fidelity, the proposed ICF-driven framework serves as a robust preprocessing stage for ophthalmic image analysis. Its integration can improve the accuracy and reliability of downstream tasks such as vessel segmentation, optic disc detection, and lesion characterization, thereby supporting both automated systems and clinical decision-making in ophthalmic diagnosis.