2020/11/27 by Mu Cai, Hong Zhang, Cai, Mu +9 · 7 citations
Computer Science · Engineering · #Acoustics #Advanced Vision and Imaging #Artificial intelligence #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Computer vision #FOS: Computer and information sciences #Feature (linguistics) #Frequency domain #Generative Adversarial Networks and Image Synthesis #Identity (music) #Image (mathematics) #Image Processing Techniques and Applications #Image quality #Image translation #Pattern recognition (psychology) #Translation (biology) #cs.CV
paper · pdf · doi:10.48550/arxiv.2011.13611
15 pages
openalex publication_date 2020/11/27 · arxiv created 2021/08/05 · arxiv updated 2021/08/06 · openalex created_date 2021/10/11 · openalex updated_date 2026/07/28
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.