D-CGRet: a multistage hybrid framework for denoising and contrast improvement in the retinal fundus image

Introduction Retinal fundus image enhancement is an important step to improve the visibility of retinal structures and support automated ophthalmic image analysis. In this study, we propose D-CGRet, a hybrid retinal fundus image enhancement framework that combines filter-based denoising, Contrast Limited Adaptive Histogram Equalization (CLAHE), Adaptive Gamma Correction with Weighting Distribution (AGCWD), and Multi-Scale Retinex (MSR). Methods The proposed framework employs a sequential enhancement pipeline to address noise degradation, local contrast limitations, and non-uniform illumination while preserving the relevant retinal information. In the first experiment, 45 noise-augmented test images were used to investigate the framework’s performance under Gaussian, salt-and-pepper, and speckle noise conditions. Four denoising filters (median, Gaussian, Wiener, and average filters) were evaluated independently, followed by CLAHE on the value component in the hue, saturation, and value (HSV) color space, adaptive gamma correction and MSR- based illumination enhancement. In additional experiments, the selected original retinal fundus images from the Fundus Photo Enhancement (FPE) dataset and the Digital Retinal Images for Vessel Extraction (DRIVE) dataset were assessed for various image-quality conditions, including poor-quality, normal, over-contrasted, and dark images. Results and discussion The evaluation was performed through comparison with existing enhancement techniques, the visual inspection of the enhanced outputs, and the retinal blood-vessel analysis of the original and processed images. The performance was evaluated using Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), Correlation Coefficient (CoC), Edge Preservation Index (EPI), Contrast Improvement Index (CII), and entropy. The proposed framework achieved SSIM values of 0.94 and 0.92 for the normal image and the dark image, respectively, indicating high structural similarity under the image-quality conditions. The experimental contributions combine to show a systematic evaluation of noise suppression, contrast enhancement, illumination correction, and preservation of vascular details. Although the results show the potential of D-CGRet to improve retinal fundus images for the cases studied, the small number of different original images means that the findings need to be further validated on larger and more diverse datasets for generalizability.

Authors

Institutions

Publication Details

Journal
Frontiers in Ophthalmology
Published
2026-11-18
DOI
https://doi.org/10.3389/fopht.2026.1946223
Primary Topic
Image Enhancement Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

D-CGRet: a multistage hybrid framework for denoising and contrast improvement in the retinal fundus image

Shivani Gupta, Suman Jaiswar
Frontiers in Ophthalmology
Image Enhancement Techniques
article

D-CGRet: a multistage hybrid framework for denoising and contrast improvement in the retinal fundus image

Shivani Gupta, Suman Jaiswar
article en

Abstract

Introduction Retinal fundus image enhancement is an important step to improve the visibility of retinal structures and support automated ophthalmic image analysis. In this study, we propose D-CGRet, a hybrid retinal fundus image enhancement framework that combines filter-based denoising, Contrast Limited Adaptive Histogram Equalization (CLAHE), Adaptive Gamma Correction with Weighting Distribution (AGCWD), and Multi-Scale Retinex (MSR). Methods The proposed framework employs a sequential enhancement pipeline to address noise degradation, local contrast limitations, and non-uniform illumination while preserving the relevant retinal information. In the first experiment, 45 noise-augmented test images were used to investigate the framework’s performance under Gaussian, salt-and-pepper, and speckle noise conditions. Four denoising filters (median, Gaussian, Wiener, and average filters) were evaluated independently, followed by CLAHE on the value component in the hue, saturation, and value (HSV) color space, adaptive gamma correction and MSR- based illumination enhancement. In additional experiments, the selected original retinal fundus images from the Fundus Photo Enhancement (FPE) dataset and the Digital Retinal Images for Vessel Extraction (DRIVE) dataset were assessed for various image-quality conditions, including poor-quality, normal, over-contrasted, and dark images. Results and discussion The evaluation was performed through comparison with existing enhancement techniques, the visual inspection of the enhanced outputs, and the retinal blood-vessel analysis of the original and processed images. The performance was evaluated using Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), Correlation Coefficient (CoC), Edge Preservation Index (EPI), Contrast Improvement Index (CII), and entropy. The proposed framework achieved SSIM values of 0.94 and 0.92 for the normal image and the dark image, respectively, indicating high structural similarity under the image-quality conditions. The experimental contributions combine to show a systematic evaluation of noise suppression, contrast enhancement, illumination correction, and preservation of vascular details. Although the results show the potential of D-CGRet to improve retinal fundus images for the cases studied, the small number of different original images means that the findings need to be further validated on larger and more diverse datasets for generalizability.

Frontiers in OphthalmologyVol. 6
Vellore Institute of Technology University (IN)
Good health and well-being
Image Enhancement Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon