Skip to Main content Skip to Navigation
Conference papers

Binary Probability Model for Learning Based Image Compression

Abstract : In this paper, we propose to enhance learned image compression systems with a richer probability model for the latent variables. Previous works model the latents with a Gaussian or a Laplace distribution. Inspired by binary arithmetic coding , we propose to signal the latents with three binary values and one integer, with different probability models. A relaxation method is designed to perform gradient-based training. The richer probability model results in a better entropy coding leading to lower rate. Experiments under the Challenge on Learned Image Compression (CLIC) test conditions demonstrate that this method achieves 18 % rate saving compared to Gaussian or Laplace models.
Complete list of metadatas

https://hal-univ-rennes1.archives-ouvertes.fr/hal-02476067
Contributor : Théo Ladune <>
Submitted on : Thursday, February 20, 2020 - 2:31:08 PM
Last modification on : Wednesday, October 14, 2020 - 3:53:05 AM
Long-term archiving on: : Thursday, May 21, 2020 - 12:34:15 PM

Files

binarisation_article.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02476067, version 1
  • ARXIV : 2002.09259

Citation

Théo Ladune, Pierrick Philippe, Wassim Hamidouche, Lu Zhang, Olivier Deforges. Binary Probability Model for Learning Based Image Compression. ICASSP (International Conference on Acoustics, Speech, and Signal Processing) 2020, IEEE, May 2020, Barcelone, Spain. ⟨hal-02476067⟩

Share

Metrics

Record views

124

Files downloads

173