< Explain other AI papers

Steering One-Step Diffusion Model with Fidelity-Rich Decoder for Fast Image Compression

Zheng Chen, Mingde Zhou, Jinpei Guo, Jiale Yuan, Yifei Ji, Yulun Zhang

2025-08-08

Steering One-Step Diffusion Model with Fidelity-Rich Decoder for Fast
  Image Compression

Summary

This paper talks about SODEC, a new fast image compression model that uses a single-step diffusion process combined with advanced decoding methods to keep images looking good while making file sizes smaller.

What's the problem?

The problem is that current image compression methods often struggle to balance speed and quality, either compressing images slowly or losing important details during compression.

What's the solution?

The solution was to create SODEC, which uses a pre-trained Variational Autoencoder (VAE) to help decode images quickly, along with a fidelity guidance module that ensures the compressed images maintain high visual quality.

Why it matters?

This matters because it allows for faster and better image compression, making it easier to save and share high-quality images efficiently without long waits or big files.

Abstract

SODEC, a single-step diffusion image compression model, enhances decoding speed and fidelity by using a pre-trained VAE and a fidelity guidance module.