Guardado en:
Detalles Bibliográficos
Autores principales: Ma, Lichen, Guo, Zipeng, He, Yu, Fu, Xiaolong, Liu, Luohang, Fu, Jingling, Huang, Junshi, Li, Yan
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:https://arxiv.org/abs/2605.17759
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914576199581696
author Ma, Lichen
Guo, Zipeng
He, Yu
Fu, Xiaolong
Liu, Luohang
Fu, Jingling
Huang, Junshi
Li, Yan
author_facet Ma, Lichen
Guo, Zipeng
He, Yu
Fu, Xiaolong
Liu, Luohang
Fu, Jingling
Huang, Junshi
Li, Yan
contents To circumvent the inherent fidelity bottlenecks and optimization misalignment of VAE-based latent diffusion, pixel-space diffusion models have emerged as a compelling end-to-end paradigm. However, existing pixel diffusion models often struggle to balance computational efficiency with the preservation of high-frequency details. They frequently resort to patch-based compression or restricted local decoding, leading to a "spectral compromise" where high-frequency and fine-grained pixel information are suppressed. To address these challenges, we propose \textbf{FrequencyBooster}, a novel framework designed to empower pixel diffusion with full-frequency modeling capabilities without prohibitive overhead. The core of our method is a high-capacity decoder that specializes in extracting exhaustive high-frequency details and low-frequency semantics, the latter of which is derived from a Diffusion Transformer (DiT) backbone. Unlike prior works that sacrifice global context for local refinement, FrequencyBooster leverages high-dimensional feature representations to maintain global structural integrity while achieving superior pixel-level precision. Extensive experiments on ImageNet demonstrate the effectiveness of our approach: our model achieves a state-of-the-art FID of \textbf{1.60} at $256 \times 256$ resolution within only 320 epochs. Furthermore, at $512 \times 512$ resolution, FrequencyBooster attains an FID of \textbf{1.69}, significantly outperforming existing pixel-space and latent-space generative models.
format Preprint
id arxiv_https___arxiv_org_abs_2605_17759
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle FrequencyBooster: Full-Frequency Modeling for High-Fidelity Pixel Diffusion
Ma, Lichen
Guo, Zipeng
He, Yu
Fu, Xiaolong
Liu, Luohang
Fu, Jingling
Huang, Junshi
Li, Yan
Computer Vision and Pattern Recognition
To circumvent the inherent fidelity bottlenecks and optimization misalignment of VAE-based latent diffusion, pixel-space diffusion models have emerged as a compelling end-to-end paradigm. However, existing pixel diffusion models often struggle to balance computational efficiency with the preservation of high-frequency details. They frequently resort to patch-based compression or restricted local decoding, leading to a "spectral compromise" where high-frequency and fine-grained pixel information are suppressed. To address these challenges, we propose \textbf{FrequencyBooster}, a novel framework designed to empower pixel diffusion with full-frequency modeling capabilities without prohibitive overhead. The core of our method is a high-capacity decoder that specializes in extracting exhaustive high-frequency details and low-frequency semantics, the latter of which is derived from a Diffusion Transformer (DiT) backbone. Unlike prior works that sacrifice global context for local refinement, FrequencyBooster leverages high-dimensional feature representations to maintain global structural integrity while achieving superior pixel-level precision. Extensive experiments on ImageNet demonstrate the effectiveness of our approach: our model achieves a state-of-the-art FID of \textbf{1.60} at $256 \times 256$ resolution within only 320 epochs. Furthermore, at $512 \times 512$ resolution, FrequencyBooster attains an FID of \textbf{1.69}, significantly outperforming existing pixel-space and latent-space generative models.
title FrequencyBooster: Full-Frequency Modeling for High-Fidelity Pixel Diffusion
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2605.17759