Practical Full Resolution Learned Lossless Image Compression

By Fabian Mentzer, Eirikur Agustsson, Michael Tschannen, Radu Timofte, Luc Van Gool. 2018

248 citations recorded by the source corpus.

We propose the first practical learned lossless image compression system, L3C, and show that it outperforms the popular engineered codecs, PNG, WebP and JPEG 2000. At the core of our method is a fully parallelizable hierarchical probabilistic model for adaptive entropy coding which is optimized end-to-end for the compression task. In contrast to recent autoregressive discrete probabilistic models such as PixelCNN, our method i) models the image distribution jointly with learned auxiliary representations instead of exclusively modeling the image distribution in RGB space, and ii) only requires three forward-passes to predict all pixel probabilities instead of one for each pixel. As a result, L3C obtains over two orders of magnitude speedups when sampling compared to the fastest PixelCNN variant (Multiscale-PixelCNN). Furthermore, we find that learning the auxiliary representation is crucial and outperforms predefined auxiliary representations such as an RGB pyramid significantly.

Subfields: Efficiency & Compression

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Efficiency & Compression

1. Pieter Abbeel
2. Jonathan Ho
3. Zhenguo Li
4. Max Welling
5. Mostafa Dehghani
6. Evan Lohn
7. Ning Kang
8. Shifeng Zhang

Citations in the corpus

• Generating Diverse High-Fidelity Images with VQ-VAE-2 (2019, methodology): “L3C [23] is a parallel architecture proposed for lossless image compression that uses jointly learned with auxiliary latent spaces to achieve speedups in sampling compared to autoregressive models.”
• iVPF: Numerical Invertible Volume Preserving Flow for Efficient Lossless Compression (2021, methodology): “L3C [26] augments the data variable with hierarchical latents, and the compression ratio is compromised due to the introduction of latent variables.”
• Compression with Flows via Local Bits-Back Coding (2019, background): “There is already a rich literature of tailored coding algorithms for autoregressive models and variational autoencoders built from conditional distributions which are already tractable for coding [38, 12, 18, 13, 42], but there are currently no such algorithms for general types of flow models [32].”
• iFlow: Numerically Invertible Flows for Efficient Lossless Compression via a Uniform Coder (2021, background): “These include, L3C [30], Bit-swap [26], HilLoc [37], and flow-based models IDF [18], IDF++ [4], iVPF [41], LBB [17].”
• IDF++: Analyzing and Improving Integer Discrete Flows for Lossless Compression (2020, background): “Mentzer et al. [33] encode an image into a set of discrete multiscale latent vectors that can be stored efficiently.”
• Integer Discrete Flows and Lossless Compression (2019, methodology): “They are used for lossless compression with deterministic encoders [25] and through bits-back coding.”