Acceleration Method for Super-Resolution Based on Diffusion Models by Intermediate Step Prediction

Jichen Ma*, Hiroshi Watanabe

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this paper, we propose a new method to improve the generation speed of single-image super-resolution models based on the diffusion model. We address the generation speed problem of super-resolution models based on the diffusion model and propose an acceleration method by predicting intermediate steps. The proposed method is highly compatible with other sampling acceleration methods while maintaining high image quality and improving the efficiency and quality of the super-resolution task.

Original languageEnglish
Title of host publicationGCCE 2024 - 2024 IEEE 13th Global Conference on Consumer Electronics
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages162-163
Number of pages2
ISBN (Electronic)9798350355079
DOIs
Publication statusPublished - 2024
Event13th IEEE Global Conference on Consumer Electronic, GCCE 2024 - Kitakyushu, Japan
Duration: 2024 Oct 292024 Nov 1

Publication series

NameGCCE 2024 - 2024 IEEE 13th Global Conference on Consumer Electronics

Conference

Conference13th IEEE Global Conference on Consumer Electronic, GCCE 2024
Country/TerritoryJapan
CityKitakyushu
Period24/10/2924/11/1

Keywords

  • clip
  • diffusion model
  • super-resolution
  • text-to-image
  • variational autoencoders

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Vision and Pattern Recognition
  • Human-Computer Interaction
  • Signal Processing
  • Electrical and Electronic Engineering
  • Media Technology
  • Instrumentation

Fingerprint

Dive into the research topics of 'Acceleration Method for Super-Resolution Based on Diffusion Models by Intermediate Step Prediction'. Together they form a unique fingerprint.

Cite this