Lossless Image Compression by Joint Prediction of Pixel and Context Using Duplex Neural Networks

This paper presents a new lossless image compression method based on the learning of pixel values and contexts through multilayer perceptrons (MLPs). The prediction errors obtained by MLPs are forwarded to adaptive arithmetic encoders, like conventional schemes. MLP-based has long been attempted for compression, recently convolutional neural networks (CNNs) also adopted lossy/lossless coding. While existing schemes focused only accurate prediction, we jointly predict contexts. We adopt design channel-wise progressive learning, residual duplex network in this framework, which leads improved coding gain compared methods. Experiments show that proposed performs better than non-learning algorithms recent learning-based methods with practical computation time.