PulseAugur
实时 10:12:39

Researchers improve image denoising with data-centric training and self-ensemble

Researchers have developed a new approach for Gaussian color image denoising, focusing on data-centric training and self-ensemble techniques rather than novel model architectures. By expanding the training dataset with larger and more diverse public image corpora and implementing a two-stage optimization process, they significantly improved performance. The method achieved a PSNR of 30.762 dB and SSIM of 0.861 on the NTIRE 2026 challenge validation set, outperforming the baseline Restormer model by over 3 dB. AI

影响 Improves image denoising techniques by emphasizing data and ensemble methods over new model architectures.

排序理由 This is a research paper detailing a new method for image denoising submitted to a challenge.

在 arXiv cs.CV 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →

Researchers improve image denoising with data-centric training and self-ensemble

报道来源 [1]

  1. arXiv cs.CV TIER_1 English(EN) · Gengjia Chang, Xining Ge, Weijun Yuan, Zhan Li, Qiurong Song, Luen Zhu, Shuhong Liu ·

    Beyond Model Design: Data-Centric Training and Self-Ensemble for Gaussian Color Image Denoising

    arXiv:2604.11468v2 Announce Type: replace Abstract: This paper presents our solution to the NTIRE 2026 Image Denoising Challenge (Gaussian color image denoising at fixed noise level $\sigma = 50$). Rather than proposing a new restoration backbone, we revisit the performance bound…