One Step Diffusion-Based Super-Resolution With Time-Aware Distillation

doi:10.1109/TIP.2026.3672376

2026-03-16

PRE

DOI：10.1109/TIP.2026.3672376

原文链接

原文求助

摘要

En 中文

基于扩散的图像超分辨率（SR）在从低分辨率输入中恢复高保真细节方面展现了巨大潜力。然而，其需要数十或数百步采样，导致显著的推理延迟。近期研究尝试通过知识蒸馏来加速这一过程，但通常仅依赖像素级损失，或忽视了扩散模型在不同时间步捕获不同信息的特性。为解决此问题，我们提出TAD-SR，一种时间感知的扩散蒸馏框架。具体而言，我们引入一种新颖的分数蒸馏策略，在微小噪声扰动后对齐学生模型与教师模型输出的分数函数。该蒸馏策略消除了分数蒸馏采样（SDS）中的固有偏差，并通过在小时间步采样，使学生模型更专注于高频图像细节。此外，我们引入时间感知判别器，利用教师知识结合显式的时间条件，在不同噪声尺度上区分真实与合成样本。在超分辨率任务上的大量实验表明，TAD-SR优于现有的单步扩散方法，并达到了与多步最先进模型相当的性能。

Keyword:

Accelerating diffusion model

super-resolution

single-step sampling

face restoration

期刊

IEEE Transactions on Image Processing

IF:

13.7

论文数:

1.0W

被引数:

8.4W

机构

huawei

学者数:

论文数: 15

被引数: 0

xidian university

学者数:

6.6K

论文数: 2.2K

被引数: 0

引用论文

RestoreFormer: High-Quality Blind Face Restoration from Undegraded Key-Value Pairs

2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

IF0

2022-06-01

OAAI

Zhouxia Wang; Jiawei Zhang; Runjian Chen; Wenping Wang; Ping Luo

Image Quality Assessment: Unifying Structure and Texture Similarity

IEEE Transactions on Pattern Analysis and Machine Intelligence

IF0

2020-01-01

OAAI

Keyan Ding; Kede Ma; Shiqi Wang; Eero P. Simoncelli

Toward Real-World Single Image Super-Resolution: A New Benchmark and a New Model

2019 IEEE/CVF International Conference on Computer Vision (ICCV)

IF0

2019-10-01

OAAI

Jianrui Cai; Hui Zeng; Hongwei Yong; Zisheng Cao; Lei Zhang

NTIRE 2017 Challenge on Single Image Super-Resolution: Methods and Results

2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

IF0

2017-07-01

OAAI

Radu Timofte; Eirikur Agustsson; Luc Van Gool; Ming-Hsuan Yang; Lei Zhang; Bee Lim; Sanghyun Son; Heewon Kim; Seungjun Nah; Kyoung Mu Lee; Xintao Wang; Yapeng Tian; Ke Yu; Yulun Zhang; Shixiang Wu; Chao Dong; Liang Lin; Yu Qiao; Chen Change Loy; Woong Bae; Jaejun Yoo; Yoseob Han; Jong Chul Ye; Jae-Seok Choi; Munchurl Kim; Yuchen Fan; Jiahui Yu; Wei Han; Ding Liu; Haichao Yu; Zhangyang Wang; Honghui Shi; Xinchao Wang; Thomas S. Huang; Yunjin Chen; Kai Zhang; Wangmeng Zuo; Zhimin Tang; Linkai Luo; Shaohui Li; Min Fu; Lei Cao; Wen Heng; Giang Bui; Truc Le; Ye Duan; Dacheng Tao; Ruxin Wang; Xu Lin; Jianxin Pang; Jinchang Xu; Yu Zhao; Xiangyu Xu; Jinshan Pan; Deqing Sun; Yujin Zhang; Xibin Song; Yuchao Dai; Xueying Qin; Xuan-Phung Huynh; Tiantong Guo; Hojjat Seyed Mousavi; Tiep Huu Vu; Vishal Monga; Cristovao Cruz; Karen Egiazarian; Vladimir Katkovnik; Rakesh Mehta; Arnav Kumar Jain; Abhinav Agarwalla; Ch V. Sai Praveen; Ruofan Zhou; Hongdiao Wen; Che Zhu; Zhiqiang Xia; Zhengtao Wang; Qi Guo

Towards Real-World Blind Face Restoration with Generative Facial Prior

2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

IF0

2021-06-01

OAAI

Xintao Wang; Yu Li; Honglun Zhang; Ying Shan

High-Resolution Image Synthesis with Latent Diffusion Models 具有潜在扩散模型的高分辨率图像合成

2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

IF0

2022-06-01

OAAI

Robin Rombach; Andreas Blattmann; Dominik Lorenz; Patrick Esser; Bjorn Ommer