Image Compression Research Based on Convolutional Autoencoder

The article presents the results of the study on image compression algorithms based on neural networks. The study analyses classical compression methods, such as JPEG, PNG, GIF, TIFF and identifies the advantages of neural network methods, in particular the use of an autoencoder, a variational autoencoder, and generative adversarial networks. A comparative analysis of classical compression algorithms, such as JPEG, with new approaches based on neural networks is carried out using the example of an autoencoder. A mathematical model describing the principle of operation for an autoencoder is presented, illustrating how a neural network encodes and restores images using latent space. To achieve the best reconstruction quality, a hybrid loss function comprising three components was employed: perceptual loss based on VGG16, SSIM loss, and MSE loss. A modular software system was developed using the Python programming language to conduct the experiments. The software includes a graphical interface, a compression module for encoding and decoding images using an autoencoder model, and a quality assessment module for calculating the main quality. The study found that traditional image compression methods demonstrate high efficiency, but are more prone to generating artifacts, especially at high compression levels, compared to neural network methods. The research results indicate that the autoencoder model can encode and decode images with minimal loss of quality, on par with JPEG, but is inferior to classical algorithms in speed and compression ratio.