The Use of Autoencoders for Noisy Images Restoration

This work is devoted to solving the problem of restoring noised images. It is proposed to use convolutional autoencoders as a solution. The paper investigates the architecture of a convolutional autoencoder based on Residual blocks using a discriminator network. The effectiveness of the models was evaluated on artificially noisy color images. Gaussian and Speckle noise models were used to artificially distort the images. Subjective and objective metrics were used to evaluate the quality of the restored images. In the course of a comparative study, conclusions were drawn about the effectiveness of using autoencoders in restoring the quality of distorted images.