An approach for de-noising and contrast enhancement of retinal fundus image using CLAHE

doi:10.1016/j.optlastec.2018.06.061

Optics & Laser Technology

Volume 110, February 2019, Pages 87-98

https://doi.org/10.1016/j.optlastec.2018.06.061 Get rights and content

Highlights

•
Proposed a noise removal and contrast enhancement algorithm for color fundus image.
•
Our method uses integration of filters and contrast limited adaptive histogram equalization (CLAHE) technique.
•
Efficacy of the method is evaluated through different performance parameters.
•
The performance of our method is better than other state-of-the-art technique.

Abstract

Now-a-days medical fundus images are widely used in clinical diagnosis for the detection of retinal disorders. Fundus images are generally degraded by noise and suffer from low contrast issues. These issues make it difficult for ophthalmologist to detect and interpret diseases in fundus images. This paper presents a noise removal and contrast enhancement algorithm for fundus image. Integration of filters and contrast limited adaptive histogram equalization (CLAHE) technique is applied for solving the issues of de-noising and enhancement of color fundus image. The efficacy of the proposed method is evaluated through different performance parameters like Peak Signal to Noise Ratio (PSNR), Structural Similarity Index (SSIM), Correlation coefficient (CoC) and Edge preservation index (EPI). The proposed method achieved 7.85% improvement in PSNR, 1.19% improvement in SSIM, 0.12% improvement in CoC and 1.28% improvement in EPI when compared to the state of the art method.

Introduction

Digital images cover a vast area of application and various applications are implemented in the processing of digital images [23], [24], [25]. Medical image processing involves input of medical images, processing and output of result. Preprocessing of medical images is carried out at the very beginning before execution of processing step in order to enhance the image by denoising and contrast enhancement [1], [2], [9], [18], [19]. Now-a-days, medical image enhancement for better diagnosis of diseases is one of the uplifting areas of interest among researchers and physicians. The main aim of medical image enhancement is reduction of noise level and improvement of the contrast of medical images [3]. One of the featured and important medical images is fundus image (i.e. retinal scan). Fundus camera is used to image digital fundus images which are used for retrieving the features like the retina, optic disc area and cup area, the posterior surface of an eye and macular regions. Digital Fundus images are widely applied for the detection of the multiple disorders related to eye [5], [7], [8], [10], [12], [14]. Abnormal eye conditions like age-related macular degeneration (AMD), glaucoma, diabetic retinopathy (DR) and neovascularization are diagnosed using fundus image [20]. Fundus image is captured using standard modality of imaging i.e. Fundus camera, which is popularly used in hospitals and eye specialist clinics. Fig. 1 shows the retinal fundus image. It is useful in the extraction of some essential features of the retina like Optic Disk Area (ODA), Cup Area Fovea, exudates and mainly the blood vessels. Some of the disorders that can be diagnosed by these features analysis in fundus images are Diabetic Retinopathy [7], [12], glaucoma [10], hypertension, etc.

Noise in fundus image can be acquired due to multiple causes like type of image modality used to capture fundus image, image acquisition procedure through fundus camera, transmission also cause noisy pixels to occur in fundus image and uneven illumination is also a key factor for presence of noise in the fundus image. Different undesirable patterns also result noise in fundus image and are caused due to following reasons [16]:

(a)
Digitization process causes low intensity white noise
(b)
Noise may occur due to different linear and non-linear patterns like large or small bright and dark irregular areas.

Fundus image is mainly affected by Gaussian (white) noise and salt and pepper (impulse) noise. In order to properly diagnose the fundus image, pre-processing or image enhancement is a key step. Quality of fundus image is a key aspect in order to execute proper diagnosis of the disease by the ophthalmologist (with the help of fundus image). Presence of noise effects the two main aspects of the fundus image i.e. sensitivity (Probability measure which specify whether fundus image classified (by any proposed model) abnormal is abnormal in real) and specificity (Probability measure which specify whether fundus image classified (by any proposed model) normal is normal in real) [10].

Preprocessing step in fundus image processing is a crucial step in order to detect and remove noise from it [3]. Feature extraction from fundus image like fovea, blood vessels, optical disk [8] and normalization of fundus images are employed for the convenience of the doctors to efficiently detect abnormalities in the eye. Fundus image is RGB in nature. Green component of fundus image is the most important as it provides most of the feature extraction while blue component is the least important one. Fundus RGB image can be decomposed into red (R), green (G) and blue (B) components as shown in Fig. 2 and then essential features can be extracted from them [6], separate pre-processing can be done on them, and separate de-noising can be performed on these segments. The most important channel is Green channel as most of the important features can be extracted through it [5]. Fig. 3 shows the diagnostic features that can be extracted from individual components of fundus image.

The rest of the paper is organized as follows: Section 2 discusses about the standard filters used to remove the noise from fundus image. Related noises associated with the fundus image are discussed in this section. This section also reviews the current work in the area of fundus image enhancement and feature extraction from it. Major contributions of proposed work towards the fundus image enhancement are also discussed in this section. The proposed model and algorithm are given in Section 3. The results and analysis of the work are discussed in Section 4. Conclusion of the proposed work is given in Section 5.

Section snippets

Background

This paper mainly focuses on the enhancement of digital fundus image by removing noise from all the three channels of the image (R, G and B) and improving its low contrast with CLAHE technique [4]. Different filters like median, Gaussian, wiener, average and weighted median are used separately for the purpose of de-noising of fundus image.

Proposed model and algorithm

The proposed model uses different filters along with CLAHE technique to remove Gaussian and salt and pepper noise and enhance the red, green and blue channels of fundus image. The combined features of CLAHE and filtering approach enhances and removes the affected noise from the R, G and B channels thus results contrast enhanced and de-noised fundus image. Fig. 6 shows the detailed diagram of proposed method. The details of algorithmic steps of the proposed method are given below:

STEP 1: Reading

Experimental results and discussion

RGB fundus image of size (605 × 700) which is collected from the STARE database [15] is used for simulation of the proposed technique. MATLAB R2016a is the tool used to simulate the proposed algorithm. Fig. 7 shows the fundus image enhancement, going through each step of the proposed algorithm with salt and pepper noise, at 0.01 noise variance level and de-noised using median filter with CLAHE technique. Similarly simulation is done for Gaussian noise found in fundus image against all types of

Conclusion and future directions

A new medical fundus image enhancement algorithm was proposed in this paper. Fundus RGB image was first decomposed into its individual red, green and blue channels and then different filtering techniques were applied along with CLAHE to de-noise and improve contrast of the image. At last, components were merged together to form enhanced RGB fundus image. It is proved that the proposed technique removes noise and enhances contrast in fundus images effectively. Various performance and quality

Acknowledgements

The authors thank the anonymous reviewers for helpful and constructive comments that greatly contributed to improving this article.

References (28)

S.M. Pizer et al.
Adaptive histogram equalization and its variations
Comput. Vis.Graph. Image Process.
(1987)
N.M. Salem et al.
Novel and adaptive contribution of the red channel in pre-processing of colour fundus images
J. Franklin Inst.
(2007)
M. Sundaram et al.
Histogram Modified Local Contrast Enhancement for mammogram images
Applied soft computing.
(2011)
S. Zhang et al.
Plant diseased leaf segmentation and recognition by fusion of superpixel, K-means and PHOG
Optik-Int. J. Light Electron Opt.
(2018)
S. Zhang et al.
Combining modified LBP and weighted SRC for palmprint recognition
Signal Image Video Process.
(2018)
S. Serikawa et al.
Underwater image dehazing using joint trilateral filter
Comput. Electric. Eng.
(2014)
H. Lu et al.
Brain intelligence: go beyond artificial intelligence
Mob. Networks Appl.
(2017)
R.C. Gonzales et al.
Digital Image Processing
(2002)
A.K. Jain
Fundamentals of Digital Image Processing
(1989)
N. Goel, A. Yadav, B.M. Singh, November. Medical image processing: a review, in: 2016 Second International Innovative...

A.W. Setiawan, T.R. Mengko, O.S. Santoso, A.B. Suksmono, Color retinal image enhancement using CLAHE, in: 2013...

H. Ab Rahim, A.S. Ibrahim, W.M.D.W. Zaki, A. Hussain, Methods to enhance digital fundus image for diabetic retinopathy...

K. Noronha, J. Nayak, S.N. Bhat, Enhancement of retinal fundus image to highlight the features for detection of...

A.K. Boyat, B.K. Joshi, A review paper: noise models in digital image processing, 2015, available form: <arXiv preprint...

Cited by (221)

A novel bronchoscopic video enhancement and tissue segmentation method based on Eulerian video magnification
2024, Biomedical Signal Processing and Control
A crucial tool for the early diagnosis of serious respiratory illness problems including lung cancer is bronchoscopy. However, the bronchoscopic video acquisition environment is constrained and unevenly illuminated, making it difficult for doctors to precisely identify intact human blood vessels and other tissue features with the naked eye. It has been demonstrated that using Eulerian video magnification techniques significantly enhances subtle changes in the video that are hardly perceptible to the human eye. We propose a novel Eulerian video linear enhancement framework for bronchoscopic video enhancement. Bronchoscopy videos are recorded and processed using Laplacian pyramid decomposition and temporal filtering algorithms to obtain the frequency bands of interest, the magnification factors of the respective bands are obtained by the constructed gain curve equations, and the magnified sub-bands are subjected to correlated noise denoising by block-matching and 3D filtering methods with exact transform-domain variance, finally we reconstruct the video. A novel framework for vessel recognition segmentation of bronchoscopic videos is proposed, thus realizing an end-to-end processing program to meet the needs of respiratory physicians. The proposed method enables remarkable enhance tissue features in bronchoscopic video and has a good effect of suppressing noise and artifacts generated during the enhancement. Tiny vessels that were previously challenging to see using the naked eye could currently be observed, and the pulsing signal of blood vessels is significantly enhanced. The experimental results demonstrate that our method is superior to the earlier method in terms of visual quality, quantitative evaluation and subjective evaluation.
Label-free microscopic cell images adaptive enhancement via weighted fusion of bright, dark, and weak structure features
2024, Biomedical Signal Processing and Control
Label-free microscopic imaging techniques have been essential for cell biology research and clinical diagnosis. However, due to low contrast, weak edges, and blurry details of label-free microscopic cell images, which bring challenges for cell analysis (detecting, segmenting, and counting, etc.). Therefore, this paper proposes an image enhancement method based on the weighted fusion of bright, dark, and weak structure features to improve the visual quality and the accuracy of cell analysis of label-free microscopic cell images. Firstly, a local sliding window analysis method is used to extract the bright and dark detail features containing background and structure features, which are weighted fusion with the input image to pre-enhance the contrast and overall brightness of images. Secondly, a feature extraction method based on guided filtering and multi-scale Gaussian filtering is proposed to extract the feature information, including the edge and structure of cells, while addressing the block artifacts. And a weighted fusion of feature information is used to obtain the weak structure feature image that has multi-scale features without a background. Finally, the weighted fusion is used to fuse the detail and edge features from the guided filtered image and the weak structure feature image to improve further details and edge features of label-free microscopic cell images and obtain the enhanced image with a uniform background while structure protecting. Experiments on PNT1A, EVICAN, and Urine datasets show that the proposed method performs superior to state-of-the-art image enhancement methods in brightness improvement, edge contrast enhancement, and detail feature preservation.
Segmentation of hard exudate lesions in color fundus image using two-stage CNN-based methods
2024, Expert Systems with Applications
The presence of hard exudate (EX) lesions is an early clinical symptom of Diabetic Retinopathy (DR); its accurate segmentation is essential for diagnosis and treatment. Automatic segmentation of EX lesions is challenging because they have a wide range of sizes, shapes, and brightness, and can be confused with soft exudate lesions. In this paper, we present and assess an efficient segmentation method for interactively segmenting EX lesions in fundus images from DR patients. Our approach consisted of two main stages: (1) the first stage generates an automatic segmentation using a CNN-based method. (2) If the obtained segmentation is suboptimal, the second stage is used to refine each inaccurately segmented region using a CNN-based interactive segmentation method. To train and evaluate method performance, we used two public datasets (IDRiD and DDR) and two local datasets from two medical centers (EHos and EWT) containing a wide range of EX lesions. Several state-of-the art CNN-based models and the proposed method were trained and evaluated on IDRiD, DDR and EHos datasets using the area under the precision–recall curve (AUPR), the dice similarity coefficient (Dice) and the intersection over union (IoU). In addition, EWT dataset was used to evaluate the methods without retraining the CNN models. The evaluation results using IDRiD dataset showed that the proposed method achieved AUPR, mean Dice and mean IoU scores of 0.893, 76.6% and 62.4%, respectively, which are comparable to the state-of-the-art method on EX segmentation. When evaluated on the DDR dataset with large portion of small EX lesions, the proposed method achieved AUPR, mean Dice and mean IoU scores of 0.693, 60.8% and 45.5%, yielding state-of-the-art performance. Furthermore, with the main evaluation metric being AUPR, the evaluation results on two datasets, EHos and ETW, with scores of 0.689 and 0.667 demonstrated superiority of the proposed method over the other segmentation methods. The proposed two-stage CNN-based segmentation method is an useful approach for accurately segmenting EX lesions in color fundus images, enabling lesion quantification to aid experts in treating patients with DR. The source code, the trained CNN models and the data are publicly available at: https://github.com/dvquang2000/Interactive_HardExudates_Segmentation.
AMDNet23: Hybrid CNN-LSTM deep learning approach with enhanced preprocessing for age-related macular degeneration (AMD) detection
2024, Intelligent Systems with Applications
In the context of a growing global population, an automated framework for disease detection can greatly assist medical professionals in diagnosing ocular diseases. This framework offers advantages such as accurate, stable, and rapid outcomes, thereby enhancing the success rate of early disease detection. The primary objective of this study was to enhance the quality of fundus images using an adaptive contrast enhancement algorithm (CLAHE) and Gamma correction as preprocessing techniques. CLAHE is employed to heighten the local contrast of fundus images, while Gamma correction enhances the intensity of relevant features. This research adopts a deep learning approach that integrates convolutional neural networks (CNNs) and both short-term and long-term memory (LSTM) mechanisms. The purpose of this combination is to automatically detect aged macular degeneration (AMD) in fundus ophthalmology. In this mechanism, CNNs are utilized to extract pertinent features from the images, and LSTM is subsequently employed to discern these extracted features. To validate the effectiveness of the proposed method, a diverse dataset of 2000 experimental fundus images is collected from various sources. These images are categorized into four distinct classes in an equitable manner. Quality assessment techniques are then applied to this dataset. The hybrid deep AMDNet23 model proposed in this study achieves successful detection of AMD ocular disease with an achieved accuracy of 96.50 %. Moreover, the system's prowess is benchmarked against 13 other pre-trained CNN models, effectively illuminating its supremacy in AMD ocular disease diagnosis within the realm of fundus imagery datasets. This exhaustive comparison underscores the method's immense potential and reaffirms its position at the forefront of cutting-edge ocular health diagnostics.
Relevance gradient descent for parameter optimization of image enhancement
2023, Computers and Graphics (Pergamon)
Machine learning plays a pivotal role in constructing learning models and leveraging abundant image data for feature extraction, which is essential for achieving accurate and efficient image enhancement through optimization algorithms. Traditional gradient descent methods are confined to optimizing independent parameters, while recent algorithms neglect the interdependence among parameters, resulting in prolonged iteration times for each method. To address this issue, we propose a novel Correlation-Based Gradient Descent (RGD) algorithm tailored for enhancing low-light images. RGD leverages inter-iteration results for mutual parameter transmission and optimization, employing convergence or oscillation criteria to select the final enhanced image and find the optimal parameter combination for image enhancement, thereby enhancing parameter optimization efficiency. Specifically, each iteration is divided into multiple operations, with each operation adopting different enhancement schemes. During each iteration, the results of different operation parameters are shared, which facilitates the discovery of optimal parameters for low-light image enhancement. Subsequently, corresponding enhancement operations are performed based on this parameter combination to optimize standard indicators. Our experiments on medical and natural low-light scenarios demonstrate that our RGD method achieves a significant reduction in iteration time, with at least a 17% decrease compared to exhaustive loop methods.
Improved bald eagle search optimization with entropy-based deep feature fusion model for breast cancer diagnosis on digital mammograms
2024, Multimedia Tools and Applications

View all citing articles on Scopus

View full text

Full length articleAn approach for de-noising and contrast enhancement of retinal fundus image using CLAHE

Highlights

Abstract

Introduction

Section snippets

Background

Proposed model and algorithm

Experimental results and discussion

Conclusion and future directions

Acknowledgements

Comput. Vis.Graph. Image Process.

J. Franklin Inst.

Applied soft computing.

Optik-Int. J. Light Electron Opt.

Signal Image Video Process.

Comput. Electric. Eng.

Mob. Networks Appl.

Digital Image Processing

Fundamentals of Digital Image Processing

Full length article
An approach for de-noising and contrast enhancement of retinal fundus image using CLAHE