Enhancing Leaf Image Quality Using Deep Learning Denoising Techniques

Abstract

It might be difficult to eliminate noise from images in the area of research. Image Denoising is a strategy used to take out distracting noise from images while preserving the original image. Plant Diseases can cause great harm to the agricultural industry. Denoising of the collected images can improve the accuracy of the identification process. Noise in images can be caused by a variety of factors such as dust, variations in light intensity, scratches, or camera artifacts. This noise can obscure the features of the plant that are used to identify the disease. Denoising can help to remove this noise and reveal the true features of the plant, which can lead to more accurate identification. There are a number of denoising approaches used to remove unwanted noise from images. In our paper, deep learning methods are used to denoise an image. We present a comparative analysis of deep image denoising approaches. In our paper, we propose four deep learning models, REDNet, MWCNN, PRIDNet, and CBDNet to identify noises in the plant leaf images. The PSNR and MSE value for these four methods is calculated and compared to the results. The greatest PSNR value is discovered to be in CBDNet. In contemporary image processing systems, image denoising is crucial. We have taken the PlantVillage dataset which comprises of 54303 diseased and normal leaf images and real images. Images of noisy leaves are given as input to the four denoising algorithms.