TY - GEN
T1 - Acceleration Method for Super-Resolution Based on Diffusion Models by Intermediate Step Prediction
AU - Ma, Jichen
AU - Watanabe, Hiroshi
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - 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.
AB - 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.
KW - clip
KW - diffusion model
KW - super-resolution
KW - text-to-image
KW - variational autoencoders
UR - http://www.scopus.com/inward/record.url?scp=85213394920&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85213394920&partnerID=8YFLogxK
U2 - 10.1109/GCCE62371.2024.10760887
DO - 10.1109/GCCE62371.2024.10760887
M3 - Conference contribution
AN - SCOPUS:85213394920
T3 - GCCE 2024 - 2024 IEEE 13th Global Conference on Consumer Electronics
SP - 162
EP - 163
BT - GCCE 2024 - 2024 IEEE 13th Global Conference on Consumer Electronics
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 13th IEEE Global Conference on Consumer Electronic, GCCE 2024
Y2 - 29 October 2024 through 1 November 2024
ER -