Lossy Image Compression with Quantized Hierarchical VAEs
arXiv:2208.13056 · doi:10.1109/WACV56688.2023.00028
Abstract
Recent research has shown a strong theoretical connection between variational autoencoders (VAEs) and the rate-distortion theory. Motivated by this, we consider the problem of lossy image compression from the perspective of generative modeling. Starting with ResNet VAEs, which are originally designed for data (image) distribution modeling, we redesign their latent variable model using a quantization-aware posterior and prior, enabling easy quantization and entropy coding at test time. Along with improved neural network architecture, we present a powerful and efficient model that outperforms previous methods on natural image lossy compression. Our model compresses images in a coarse-to-fine fashion and supports parallel encoding and decoding, leading to fast execution on GPUs. Code is available at https://github.com/duanzhiihao/lossy-vae.
arXiv commentWACV 2023 Best Algorithms Paper Award, revised version
References in corpus (12)
- Neural Discrete Representation Learning
- Variational image compression with a scale hyperprior
- NVAE: A Deep Hierarchical Variational Autoencoder
- Neural Image Compression via Non-Local Attention Optimization and Improved Context Modeling
- Non-local Attention Optimized Deep Image Compression
- CompressAI: a PyTorch library and evaluation platform for end-to-end compression research
- Asymmetric numeral systems: entropy coding combining speed of Huffman coding with compression rate of arithmetic coding
- Universally Quantized Neural Compression
- Discrete Variational Autoencoders
- DVAE#: Discrete Variational Autoencoders with Relaxed Boltzmann Priors
- Hierarchical Quantized Autoencoders
- Compressing Images by Encoding Their Latent Representations with Relative Entropy Coding