Image Noise Reduction with Autoencoder

Abstract

Digital images play an important role in our daily lives, and they can be used for a variety of applications such as fingerprint recognition, video viewing, etc. Transmission through audio channels, faults in the memory, etc. may cause noise in the image and when it is processed afterwards it will lead to inaccurate output. Therefore, there is necessity to remove the noise before in order for the image to be processed. Here comes in the need of efficient denoising technique that helps to deal with the noisy image. Image denoising is a technique to restore or repair a damaged image back to its original quality. In line with that, we sometimes need to colour the decoloured image. Image data formatting requires a special take in neural network, called Convolutional Neural Network (CNN) or Convolutional Autoencoder. For reducing the dimensions and image noise, autoencoders are widely preferred. Therefore, by its virtue, training the model to perform denoising of the images.