Quantitative analysis of Fundus Image Enhancement in the Detection of Diabetic Retinopathy Using Deep Convolutional Neural Network

Purpose: In this work, we propose a DnCNN-based method to enhance the contrast of images. Fundus images are widely used in clinical practice for detecting retinal dark spots. These spots trouble ophthalmologists to distinguish and decipher diseases for diabetic retinopathy. This work aims is to describe a fundus picture noise removal with enhancement technique. Our simulation results demonstrate that our DnCNN-based method outperforms other contrast enhancement methods. Method: This paper presents a denoising Convolutional Neutral Network (DnCNN)-based technique to enhance images for diabetic retinopathy. The contrast of images can be adaptively improved. Denoising convolutional neural networks (DnCNNs) show progress in image denoising architecture, learning algorithm, and regularization method. DnCNN removes the latent clean image in the hidden layers implicitly. Our proposed (DnCNN) method, compared with the various histogram (Equalization (HE), Adaptive Histogram Equalization (ADHE), Contrast Limited Adaptive Histogram Equalization (CLAHE), and Exposure based Sub-Image Histogram Equalization (ESIHE)) pre-processes fundus images for identifying diabetic retinopathy. Result: The fundus images were collected from the DRIVE database. The Denoising Convolutional Neural Network strategy is analyzed with the execution parameter measurements of PSNR and SSIM. According to the simulation results, the DnCNN method has a higher visual quality than the previous method. In addition, the proposed method's average PSNR values 25.4303 and 11 dB are higher than those of the existing methods. However, the DnCNN method has a slightly similar to 0.2 lower SSIM value of 0.3584 than the previous methods. The decrease in SSIM value indicates a loss of detailed information.