Forward Autoencoder Approach for Denoising Retinal OCT Images

Optical Coherence Tomography (OCT) is a noninvasive imaging method that provides high-resolution tomographic images of biological structures. Accurate medical diagnosis relies on clear images, but noise, blurriness, and inadequate denoising techniques for medical images including traditional methods like BM3D, Bilateral, Non-Local Means, Wiener, Gaussian, Average preprocessing can hinder image quality. The main goal of this work is to assess denoising the retinal OCT images using the traditional filters and proposed forward autoencoder to improve the clarity and precision of medical image data. Here, the effectiveness of these filters is analyzed using Peak Signal to Noise Ratio (PSNR), Mean Square Error (MSE), structural similarity index measure (SSIM), and Contrast to Noise Ratio (CNR). The challenges related to image copying and quality loss are addressed. From analysis, impressive results are achieved for uniform noise (PSNR: 73.771, MSE: 0.003, SSIM: 0.476, SNR: 15.23, and CNR: 4.039), Gaussian noise (PSNR: 73.460, MSE: 0.003, SSIM: 0.464, SNR: 14.905, and CNR: 3.923), Salt and pepper noise (PSNR: 73.579, MSE: 0.003, SSIM: 0.480, SNR: 15.023, and CNR: 3.955), and Speckle noise (PSNR: 72.903, MSE: 0.004, SSIM: 0.452, SNR: 14.347, and CNR: 3.676) respectively in forward autoencoder based denoising approach.