Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SEEC: Segmentation-Assisted Multi-Entropy Models for Learned Lossless Image Compression

Chunhang Zheng, Zichang Ren, Dou Li

Abstract

Recently, learned image compression has attracted considerable attention due to its superior performance over traditional methods. However, most existing approaches employ a single entropy model to estimate the probability distribution of pixel values across the entire image, which limits their ability to capture the diverse statistical characteristics of different semantic regions. To overcome this limitation, we propose Segmentation-Assisted Multi-Entropy Models for Lossless Image Compression (SEEC). Our framework utilizes semantic segmentation to guide the selection and adaptation of multiple entropy models, enabling more accurate probability distribution estimation for distinct semantic regions. Experimental results on benchmark datasets demonstrate that SEEC achieves state-of-the-art compression ratios while introducing only minimal encoding and decoding latency. With superior performance, the proposed model also supports Regions of Interest (ROIs) coding condition on the provided segmentation mask. Our code is available at https://github.com/chunbaobao/SEEC.

SEEC: Segmentation-Assisted Multi-Entropy Models for Learned Lossless Image Compression

Abstract

Recently, learned image compression has attracted considerable attention due to its superior performance over traditional methods. However, most existing approaches employ a single entropy model to estimate the probability distribution of pixel values across the entire image, which limits their ability to capture the diverse statistical characteristics of different semantic regions. To overcome this limitation, we propose Segmentation-Assisted Multi-Entropy Models for Lossless Image Compression (SEEC). Our framework utilizes semantic segmentation to guide the selection and adaptation of multiple entropy models, enabling more accurate probability distribution estimation for distinct semantic regions. Experimental results on benchmark datasets demonstrate that SEEC achieves state-of-the-art compression ratios while introducing only minimal encoding and decoding latency. With superior performance, the proposed model also supports Regions of Interest (ROIs) coding condition on the provided segmentation mask. Our code is available at https://github.com/chunbaobao/SEEC.

Paper Structure

This paper contains 19 sections, 10 equations, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Learned Lossy Image Compression
  4. Learned Lossless Image Compression
  5. Methodology
  6. Segmentation-assisted Multi-Entropy Models
  7. Multi-channel Discrete Logistic Mixture Likelihood
  8. ROIs Coding
  9. Experiments
  10. Settings
  11. Coding Performance of SEEC
  12. Efficiency of Codec
  13. ROIs Coding
  14. Ablation Studies
  15. Analysis of Network Components
  16. ...and 4 more sections

Figures (3)

  • Figure 1: Entropy Model Comparison. (a) Exiting methods apply a single entropy model to the entire image. (b) Our proposed Multi-Entropy Models assign distinct entropy models to different semantic regions, allowing for more accurate modeling of pixel value distributions.
  • Figure 2: The overall architecture of SEEC. The left part shows the Segmentation-aware Image Compressor (SIC) module, which extracts segmentation-aware features from the input image. The right part illustrates the Segmentation-assisted Multi-Entropy Models (SMEM) module, which estimates the probability distribution of pixel values using multiple entropy models guided by semantic segmentation. Q denotes the quantization operation. AE and AD represent arithmetic encoder and arithmetic decoder, respectively.
  • Figure 3: Visualization of ROIs coding. From top to bottom: original image, segmentation mask where white regions denote the regions of interest (foreground), reconstructed image using ROIs coding, and error map where brighter pixels indicate larger reconstruction errors.