The paper argues that the concrete random noise used in diffusion models is not merely a passive perturbation but a learnable input that can be exploited by the model. By analyzing how clean data and realized noise jointly form the noisy input, the authors show that the model can learn regularities in the data or in the noise structure, and that these two routes can interact. Experiments on MNIST and CIFAR‑10 using pseudorandom streams demonstrate that structured‑noise training can reduce prediction loss, but this advantage disappears when test noise is replaced with IID noise, indicating that the learned dependence is tied to the specific noise structure.
By Shengzhi Deng, Chenqi Ye, Yanze Guo
Pixel‑Space Diffusion via Observation Operators introduces a new framework for pixel‑space diffusion models that addresses a scale‑time mismatch in existing methods. By replacing fixed full‑image supervision with a time‑indexed observation trajectory that progresses from coarse structures to the full image, the model aligns supervision with the natural recovery order of image details. The approach employs Gaussian‑Lanczos operators and a GL‑CoDA decoder to refine features progressively, resulting in faster convergence and higher generation quality, achieving an FID of 1.52 on ImageNet‑256.
By Shaojie Guo, Lichen Ma, Haoyang Tong, Yu He, Zipeng Guo, Xiaoan Liu, Feng Yan, Yu Guo, Fei Wang, Junshi Huang, Yan Wang
arXiv:2608.29038v1 Announce Type: new
Abstract: Noise poses a widespread challenge in signal processing, particularly when it comes to denoising images. Although convolutional neural networks (CNNs)...
By Dongjin Kim, Donggoo Jung, Sungyong Baik, Tae Hyun Kim
arXiv:2502. 07580v4 Announce Type: replace Abstract: We present a novel view of diffusion-like generative modeling from the perspective of iterative Gaussian posterior inference.
By Marten Lienen, Marcel Kollovieh, Stephan G\"unnemann
arXiv:2610.33136v1 Announce Type: cross
Abstract: Low-count imaging arises when limited illumination or dose yields only a few detected particles in microscopy, such as photons or electrons, causing...
By Huazhen Liu, Zipei Yan, Jizhou Li
arXiv:2610.14136v2 Announce Type: new
Abstract: Iterative sampling in diffusion models can lead to error accumulation due to exposure bias. The denoiser is trained on real images interpolated with no...
By Donghoon Ahn, Runqian Wang, Alexei A. Efros, Taesung Park, Yash Belhe
arXiv:2602. 09639v2 Announce Type: replace Abstract: Denoising diffusion models (DDMs) are state-of-the-art methods for learning densities from data across numerous domains, yet many aspects of the training and sampling pipeline remain poorly understood.
By Zahra Kadkhodaie, Aram-Alexandre Pooladian, Sinho Chewi, Eero Simoncelli
The paper introduces $p$-Bridge, a unified diffusion framework for Tweedie noise that models sensor noise across Gaussian, Poisson, and Gamma regimes using a noise exponent $p$ and accumulated measurement $
u$. It defines a noise plane and a stochastic bridge that interprets denoising as progressive measurement accumulation, allowing restoration between arbitrary measurement levels. A single exponent-conditioned network learns a shared reverse-mean update and deviance-based objective, achieving cross-exponent robustness comparable to multiple specialist models and demonstrating effectiveness on synthetic images, fluorescence microscopy, and SAR data.
By Xuran Hu, Yuru Jia, Tengxi Wang, Milo\v{s} Dakovi\'{c}, Ljubi\v{s}a Stankovi\'{c}, Wufan Zhao
arXiv:2609. 31670v2 Announce Type: replace Abstract: We study rectified flows for denoising without clean training targets.
By Timothy Sereda, Debesh Jha
arXiv:2609.36210v1 Announce Type: new
Abstract: Generative denoising models, such as diffusion and flow-matching, learn to sample from complex distributions by training a deep neural network denoiser...
By Alexandros Graikos, Nebojsa Jojic, Dimitris Samaras
arXiv:2610. 19692v1 Announce Type: new Abstract: Gaussian diffusion models offer a practical framework for generative modeling.
By Yuling Jiao, Lican Kang, Jerry Zhijian Yang, Cheng Yuan, Ruizhe Zhang
The paper introduces a physics‑driven, cross‑domain iterative framework for self‑supervised low‑dose CT denoising. It first uses a learned sinogram prior and the LDCT noise model to separate Poisson and Gaussian noise components, then applies binomial and Gaussian data thinning to create two training pairs with independent noise realizations. These pairs train an image‑domain network whose outputs are forward‑projected to refine the prior, yielding consistent performance gains over existing self‑supervised baselines and comparable results to supervised methods.
By Xianlei Han, Shaoyu Wang, Jiancheng Fang, Weiwen Wu, Qiegen Liu