PulseAugur
EN
LIVE 16:58:18

YeTI framework generates realistic image noise from two noisy images

By PulseAugur Editorial · [2 sources] ·

Researchers have developed YeTI, a novel framework for generating realistic sRGB noise for image denoising tasks. This method learns to synthesize signal-dependent noise using only two noisy images of the same scene, eliminating the need for clean ground truth data or camera metadata. YeTI employs a Reconstruction Autoencoder and a Conditional Diffusion Transformer to disentangle scene structure from noise characteristics, enabling the generation of realistic noise that preserves underlying image content. Experiments show that denoisers trained with YeTI-synthesized images achieve strong real-world performance on various benchmarks, including SIDD and DND. AI

IMPACT Enables more robust image denoising models by providing a scalable method for generating realistic training data without requiring clean images.

RANK_REASON The cluster contains an academic paper detailing a new method for image noise generation.

Read on arXiv cs.CV →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

YeTI framework generates realistic image noise from two noisy images

COVERAGE [2]

  1. arXiv cs.CV TIER_1 English(EN) · Jaekyun Ko, Byung Wan Lim, Soomin Lee, Dongjin Kim, Tae Hyun Kim ·

    YeTI: You Only Need Two Noisy Images for Real-World sRGB Noise Generation

    arXiv:2607.09193v1 Announce Type: new Abstract: Real-world sRGB image denoising remains challenging due to the nonlinear characteristics of sensor noise and the difficulty of acquiring aligned clean-noisy image pairs. Supervised denoisers often overfit to limited paired datasets,…

  2. arXiv cs.CV TIER_1 English(EN) · Tae Hyun Kim ·

    YeTI: You Only Need Two Noisy Images for Real-World sRGB Noise Generation

    Real-world sRGB image denoising remains challenging due to the nonlinear characteristics of sensor noise and the difficulty of acquiring aligned clean-noisy image pairs. Supervised denoisers often overfit to limited paired datasets, while self-supervised methods still depend on s…

RELATED ENTITIES

RELATED TOPICS