Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2512.05106 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914369124696064 |
|---|---|
| author | Zeng, Yu Ochoa, Charles Zhou, Mingyuan Patel, Vishal M. Guizilini, Vitor McAllister, Rowan |
| author_facet | Zeng, Yu Ochoa, Charles Zhou, Mingyuan Patel, Vishal M. Guizilini, Vitor McAllister, Rowan |
| contents | Standard diffusion corrupts data using Gaussian noise whose Fourier coefficients have random magnitudes and random phases. While effective for unconditional or text-to-image generation, corrupting phase components destroys spatial structure, making it ill-suited for tasks requiring geometric consistency, such as re-rendering, simulation enhancement, and image-to-image translation. We introduce Phase-Preserving Diffusion (ϕ-PD), a model-agnostic reformulation of the diffusion process that preserves input phase while randomizing magnitude, enabling structure-aligned generation without architectural changes or additional parameters. We further propose Frequency-Selective Structured (FSS) noise, which provides continuous control over structural rigidity via a single frequency-cutoff parameter. ϕ-PD adds no inference-time cost and is compatible with any diffusion model for images or videos. Across photorealistic and stylized re-rendering, as well as sim-to-real enhancement for driving planners, ϕ-PD produces controllable, spatially aligned results. When applied to the CARLA simulator, ϕ-PD significantly improves sim-to-real planner transfer performance. The method is complementary to existing conditioning approaches and broadly applicable to image-to-image and video-to-video generation. Videos, additional examples, and code are available on our \href{https://yuzeng-at-tri.github.io/ppd-page/}{project page}. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_05106 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | NeuralRemaster: Phase-Preserving Diffusion for Structure-Aligned Generation Zeng, Yu Ochoa, Charles Zhou, Mingyuan Patel, Vishal M. Guizilini, Vitor McAllister, Rowan Computer Vision and Pattern Recognition Graphics Machine Learning Robotics Standard diffusion corrupts data using Gaussian noise whose Fourier coefficients have random magnitudes and random phases. While effective for unconditional or text-to-image generation, corrupting phase components destroys spatial structure, making it ill-suited for tasks requiring geometric consistency, such as re-rendering, simulation enhancement, and image-to-image translation. We introduce Phase-Preserving Diffusion (ϕ-PD), a model-agnostic reformulation of the diffusion process that preserves input phase while randomizing magnitude, enabling structure-aligned generation without architectural changes or additional parameters. We further propose Frequency-Selective Structured (FSS) noise, which provides continuous control over structural rigidity via a single frequency-cutoff parameter. ϕ-PD adds no inference-time cost and is compatible with any diffusion model for images or videos. Across photorealistic and stylized re-rendering, as well as sim-to-real enhancement for driving planners, ϕ-PD produces controllable, spatially aligned results. When applied to the CARLA simulator, ϕ-PD significantly improves sim-to-real planner transfer performance. The method is complementary to existing conditioning approaches and broadly applicable to image-to-image and video-to-video generation. Videos, additional examples, and code are available on our \href{https://yuzeng-at-tri.github.io/ppd-page/}{project page}. |
| title | NeuralRemaster: Phase-Preserving Diffusion for Structure-Aligned Generation |
| topic | Computer Vision and Pattern Recognition Graphics Machine Learning Robotics |
| url | https://arxiv.org/abs/2512.05106 |