PulseAugur
实时 20:48:12

新的扩散反演技术改进图像重建和地震分析 · 跟踪4个来源

研究人员正在开发新的扩散反演方法,该过程将图像映射回扩散模型的潜在空间以进行重建和编辑。一种方法“后验延续”(Posterior Continuation)根据噪声水平优化频带暴露,以提高各种任务的恢复性能。另一种方法“解耦潜在优化”(Decoupled Latent Optimization, DLO)通过解耦数据保真度和先验一致性来增强全波形反演,从而获得更真实的地球物理结构。此外,一种称为“时间步重排”(Timestep Rescheduling)的技术优化了噪声调度器的时间步选择,以最小化反演误差并提高现有扩散反演方法的准确性。 AI

影响 这些扩散反演技术的进步可能带来更准确、更逼真的图像重建、编辑和地下分析。

排序理由 多篇arXiv论文详细介绍了扩散反演及相关任务的新颖方法。

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 4 个来源。 我们如何撰写摘要 →

新的扩散反演技术改进图像重建和地震分析 · 跟踪4个来源

报道来源 [4]

  1. arXiv cs.AI TIER_1 English(EN) · Feng Tian, Yixuan Li, Weili Zeng, Weitian Zhang, Yichao Yan, Xiaokang Yang ·

    面向扩散逆问题的带噪声条件频率暴露的后验续接

    arXiv:2602.00176v2 Announce Type: replace-cross Abstract: Diffusion posterior sampling solves inverse problems by combining a pretrained diffusion prior with measurement-consistency guidance. However, full-band guidance can be unreliable at high noise levels, where clean estimate…

  2. arXiv cs.LG TIER_1 English(EN) · Chen Min, Zheng Ma ·

    Decoupled Latent Optimization of Diffusion Models for Full Waveform Inversion

    arXiv:2606.14139v1 Announce Type: new Abstract: Full waveform inversion (FWI) recovers subsurface velocity from seismic recordings by solving a severely ill-posed, nonconvex PDE-constrained optimization. Classical regularizers stabilize the inversion but fail to reproduce realist…

  3. arXiv cs.LG TIER_1 English(EN) · Zheng Ma ·

    Decoupled Latent Optimization of Diffusion Models for Full Waveform Inversion

    Full waveform inversion (FWI) recovers subsurface velocity from seismic recordings by solving a severely ill-posed, nonconvex PDE-constrained optimization. Classical regularizers stabilize the inversion but fail to reproduce realistic geological structures; recent diffusion-prior…

  4. arXiv cs.CV TIER_1 Deutsch(DE) · Shangquan Sun, Ting Gong, Zhirui Liu, Jiamin Wu, Runkai Zhao, Mianxin Liu, Wenqi Ren, Xiaochun Cao ·

    扩散反演中的时间步重排

    arXiv:2606.15389v1 Announce Type: new Abstract: Diffusion inversion, which maps images back to the Gaussian latent space of a diffusion model, is a critical task for image reconstruction and editing. While DDIM enables fast deterministic inversion, it inherently introduces deviat…