Open AccessDeep Autoencoder-Based Image Compression using Multi-Layer Perceptrons
Author(s) -
G.G.H.M.T.R. Bandara,
R. Siyambalapitiya
Publication year - 2020
Publication title -
international journal of soft computing and engineering
Language(s) - English
Resource type - Journals
ISSN - 2231-2307
DOI - 10.35940/ijsce.e3357.039620
Subject(s) - autoencoder , artificial neural network , artificial intelligence , computer science , deep learning , pattern recognition (psychology) , backpropagation , block (permutation group theory) , perceptron , preprocessor , image compression , pixel , normalization (sociology) , image processing , image (mathematics) , mathematics , geometry , sociology , anthropology
The Artificial Neural Network is one of the heavily used alternatives for solving complex problems in machine learning and deep learning. In this research, a deep autoencoder-based multi-layer feed-forward neural network has been proposed to achieve image compression. The proposed neural network splits down a large image into small blocks and each block applies the normalization process as the preprocessing technique. Since this is an autoencoder-based neural network, each normalized block of pixels has been initialized as the input and the output of the neural network. The training process of the proposed network has been done for various block sizes and different saving percentages of various kinds of images by using the backpropagation algorithm. The output of the middle-hidden layer will be the compressed representation for each block of the image. The proposed model has been implemented using Python, Keras, and Tensor flow backend.