Frequency Domain Image Translation: More Photo-realistic, Better Identity-preserving
Mu Cai, Hong Zhang, Huijuan Huang, Qichuan Geng, Yixuan Li, Gao Huang
Abstract
Image-to-image translation has been revolutionized with GAN-based methods. However, existing methods lack the ability to preserve the identity of the source domain. As a result, synthesized images can often over-adapt to the reference domain, losing important structural characteristics and suffering from suboptimal visual quality. To solve these challenges, we propose a novel frequency domain image translation (FDIT) framework, exploiting frequency information for enhancing the image generation process. Our key idea is to decompose the image into low-frequency and high-frequency components, where the high-frequency feature captures object structure akin to the identity. Our training objective facilitates the preservation of frequency information in both pixel space and Fourier spectral space. We broadly evaluate FDIT across five large-scale datasets and multiple tasks including image translation and GAN inversion. Extensive experiments and ablations show that FDIT effectively preserves the identity of the source image, and produces photo-realistic images. FDIT establishes state-of-the-art performance, reducing the average FID score by 5.6% compared to the previous best method.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
23 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- Swapping Autoencoder for Deep Image Manipulation cites Photorealistic Style Transfer via Wavelet Transforms
- Swapping Autoencoder for Deep Image Manipulation cites Image2StyleGAN: How to Embed Images Into the StyleGAN Latent Space?
- Swapping Autoencoder for Deep Image Manipulation cites GANSpace: Discovering Interpretable GAN Controls
- Focal Frequency Loss for Image Reconstruction and Synthesis cites Photorealistic Style Transfer via Wavelet Transforms
- AesFA: An Aesthetic Feature-Aware Arbitrary Neural Style Transfer cites Drop an Octave: Reducing Spatial Redundancy in Convolutional Neural Networks With Octave Convolution
Cited by top-tier papers23
- Focal Frequency Loss for Image Reconstruction and SynthesisLiming Jiang, Bo Dai, Wayne Wu, Chen Change LoyICCV 2021 · 422 citations
- Spectral Unsupervised Domain Adaptation for Visual RecognitionJingyi Zhang, Jiaxing Huang, Zichen Tian, Shijian LuCVPR 2022 · 72 citations
- AesFA: An Aesthetic Feature-Aware Arbitrary Neural Style TransferJoonwoo Kwon, Sooyoung Kim, Yuewei Lin, Shinjae Yoo et al.AAAI 2024 · 32 citations
- Painterly Image Harmonization in Dual DomainsJunyan Cao, Yan Hong, Li NiuAAAI 2023 · 27 citations
- Frequency-Controlled Diffusion Model for Versatile Text-Guided Image-to-Image TranslationXiang Gao, Zhengbo Xu, Junhan Zhao, Jiaying LiuAAAI 2024 · 23 citations
Builds on17
- Image2StyleGAN: How to Embed Images Into the StyleGAN Latent Space?Rameen Abdal, Yipeng Qin, Peter WonkaICCV 2019 · 1,195 citations
- GANSpace: Discovering Interpretable GAN ControlsErik Härkönen, Aaron Hertzmann, Jaakko Lehtinen, Sylvain ParisNeurIPS 2020 · 1,049 citations
- Drop an Octave: Reducing Spatial Redundancy in Convolutional Neural Networks With Octave ConvolutionYunpeng Chen, Haoqi Fan, Bing Xu, Zhicheng Yan et al.ICCV 2019 · 665 citations
- Photorealistic Style Transfer via Wavelet TransformsJaejun Yoo, Youngjung Uh, Sanghyuk Chun, Byeongkyu Kang et al.ICCV 2019 · 412 citations
- Swapping Autoencoder for Deep Image ManipulationTaesung Park, Jun-Yan Zhu, Oliver Wang, Jingwan Lu et al.NeurIPS 2020 · 376 citations
Related papers
- SEIT: Structural Enhancement for Unsupervised Image Translation in Frequency DomainZhifeng Zhu, Yaochen Li, Yifan Li, Jinhuo Yang et al.AAAI 2024 · 5 citations
- Frequency-Guided Diffusion for Training-Free Text-Driven Image TranslationZheng Gao, Jifei Song, Zhensong Zhang, Jiankang Deng et al.ICCV 2025 · 1 citation
- FSDR: Frequency Space Domain Randomization for Domain GeneralizationJiaxing Huang, Dayan Guan, Aoran Xiao, Shijian LuCVPR 2021
- Spectrum Translation for Refinement of Image Generation (STIG) Based on Contrastive Learning and Spectral Filter ProfileSeokjun Lee, Seung-Won Jung, Hyunseok SeoAAAI 2024 · 8 citations
- Wavelet Knowledge Distillation: Towards Efficient Image-to-Image TranslationLinfeng Zhang, Xin Chen, Xiaobing Tu, Pengfei Wan et al.CVPR 2022 · 105 citations
