UltraImage: Rethinking Resolution Extrapolation in Image Diffusion Transformers

• URL: http://arxiv.org/abs/2512.04504v1
• Date: Thu, 04 Dec 2025 06:24:04 GMT
• Title: UltraImage: Rethinking Resolution Extrapolation in Image Diffusion Transformers
• Abstract summary: We present UltraImage, a principled framework that addresses content repetition and quality degradation.<n>Experiments show that UltraImage consistently outperforms prior methods on Qwen-Image and Flux (around 4K) across three generation scenarios.<n>UltraImage can generate images up to 6K*6K without low-resolution guidance from a training resolution of 1328p, demonstrating its extreme extrapolation capability.
• Score: 58.37965214608629
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Recent image diffusion transformers achieve high-fidelity generation, but struggle to generate images beyond these scales, suffering from content repetition and quality degradation. In this work, we present UltraImage, a principled framework that addresses both issues. Through frequency-wise analysis of positional embeddings, we identify that repetition arises from the periodicity of the dominant frequency, whose period aligns with the training resolution. We introduce a recursive dominant frequency correction to constrain it within a single period after extrapolation. Furthermore, we find that quality degradation stems from diluted attention and thus propose entropy-guided adaptive attention concentration, which assigns higher focus factors to sharpen local attention for fine detail and lower ones to global attention patterns to preserve structural consistency. Experiments show that UltraImage consistently outperforms prior methods on Qwen-Image and Flux (around 4K) across three generation scenarios, reducing repetition and improving visual fidelity. Moreover, UltraImage can generate images up to 6K*6K without low-resolution guidance from a training resolution of 1328p, demonstrating its extreme extrapolation capability. Project page is available at \href{https://thu-ml.github.io/ultraimage.github.io/}{https://thu-ml.github.io/ultraimage.github.io/}.

Related papers

• Signal Structure-Aware Gaussian Splatting for Large-Scale Scene Reconstruction [25.32043603138213]
  We propose SIG, a novel scheduler that synchronizes image supervision with Gaussian frequencies.<n>Specifically, we derive the average sampling frequency and bandwidth of 3D representations, and then regulate the training image resolution and the Gaussian densification process.<n>Our framework enables frequency-consistent, geometry-aware, and floater-free training, achieving state-of-the-art performance by a substantial margin in both efficiency and rendering quality.
  arXiv  Detail & Related papers  (2026-07-02T04:41:51Z)
• InsightTok: Improving Text and Face Fidelity in Discrete Tokenization for Autoregressive Image Generation [67.8525902443746]
  InsightTok is a discrete visual tokenization framework that enhances text and face fidelity through localized, content-aware perceptual losses.<n>With a compact 16k codebook and a 16x downsampling rate, InsightTok significantly outperforms prior tokenizers in text and face reconstruction.<n>Results highlight the potential of specialized supervision in tokenizer training for advancing discrete image generation.
  arXiv  Detail & Related papers  (2026-05-14T03:57:25Z)
• UltraViCo: Breaking Extrapolation Limits in Video Diffusion Transformers [42.379492244355454]
  We propose a training-free, plug-and-play method that suppresses attention for tokens beyond the training window via a constant decay factor.<n>Remarkably, it improves Dynamic Degree and Imaging Quality by 233% and 40.5% over the previous best method at 4x extrapolation.
  arXiv  Detail & Related papers  (2025-11-25T09:44:10Z)
• High-Frequency First: A Two-Stage Approach for Improving Image INR [13.070432644808806]
  Implicit Neural Representations (INRs) have emerged as a powerful alternative to traditional pixel-based formats.<n>A key challenge lies in the spectral bias of neural networks, which tend to favor low-frequency components while struggling to capture high-frequency details.<n>We introduce a two-stage training strategy where a neighbor-aware soft mask adaptively assigns higher weights to pixels with strong local variations.
  arXiv  Detail & Related papers  (2025-08-21T13:56:36Z)
• GigaTok: Scaling Visual Tokenizers to 3 Billion Parameters for Autoregressive Image Generation [81.58846231702026]
  We introduce GigaTok, the first approach to improve image reconstruction, generation, and representation learning when scaling visual tokenizers.<n>We identify the growing complexity of latent space as the key factor behind the reconstruction vs. generation dilemma.<n>By scaling to $bf3 space billion$ parameters, GigaTok achieves state-of-the-art performance in reconstruction, downstream AR generation, and downstream AR representation quality.
  arXiv  Detail & Related papers  (2025-04-11T17:59:58Z)
• NFIG: Multi-Scale Autoregressive Image Generation via Frequency Ordering [47.442844594442455]
  NextFrequency Image Generation (NFIG) is a novel framework that decomposes the image generation process into multiple frequency-guided stages.<n> NFIG aligns the generation process with the natural image structure.<n>It does this by first generating low-frequency components, which efficiently capture global structure with significantly fewer tokens, and then progressively adding higher-frequency details.
  arXiv  Detail & Related papers  (2025-03-10T08:59:10Z)
• Multi-view Image Diffusion via Coordinate Noise and Fourier Attention [5.251293630298169]
  We propose a diffusion process that attends to time-dependent spatial frequencies of features with a novel attention mechanism and cross-attention loss.<n>Our technique improves SOTA on several quantitative metrics with qualitatively better results when compared to other state-of-the-art approaches for multi-view consistency.
  arXiv  Detail & Related papers  (2024-12-04T22:49:40Z)
• Hyper-VolTran: Fast and Generalizable One-Shot Image to 3D Object Structure via HyperNetworks [53.67497327319569]
  We introduce a novel neural rendering technique to solve image-to-3D from a single view. Our approach employs the signed distance function as the surface representation and incorporates generalizable priors through geometry-encoding volumes and HyperNetworks. Our experiments show the advantages of our proposed approach with consistent results and rapid generation.
  arXiv  Detail & Related papers  (2023-12-24T08:42:37Z)
• Gated Multi-Resolution Transfer Network for Burst Restoration and Enhancement [75.25451566988565]
  We propose a novel Gated Multi-Resolution Transfer Network (GMTNet) to reconstruct a spatially precise high-quality image from a burst of low-quality raw images. Detailed experimental analysis on five datasets validates our approach and sets a state-of-the-art for burst super-resolution, burst denoising, and low-light burst enhancement.
  arXiv  Detail & Related papers  (2023-04-13T17:54:00Z)
• Accurate Image Restoration with Attention Retractable Transformer [50.05204240159985]
  We propose Attention Retractable Transformer (ART) for image restoration. ART presents both dense and sparse attention modules in the network. We conduct extensive experiments on image super-resolution, denoising, and JPEG compression artifact reduction tasks.
  arXiv  Detail & Related papers  (2022-10-04T07:35:01Z)
• Wiener Guided DIP for Unsupervised Blind Image Deconvolution [10.440495513371747]
  Blind deconvolution is an ill-posed problem arising in various fields ranging from microscopy to astronomy. Deep learning architectures can serve as an image generation prior during unsupervised blind deconvolution optimization. We propose to use Wiener-deconvolution to guide the image generator during optimization by providing it a sharpened version of the blurry image.
  arXiv  Detail & Related papers  (2021-12-19T22:19:13Z)
• On Measuring and Controlling the Spectral Bias of the Deep Image Prior [63.88575598930554]
  The deep image prior has demonstrated the remarkable ability that untrained networks can address inverse imaging problems. It requires an oracle to determine when to stop the optimization as the performance degrades after reaching a peak. We study the deep image prior from a spectral bias perspective to address these problems.
  arXiv  Detail & Related papers  (2021-07-02T15:10:42Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.