Researchers have developed a new framework called ENCORE for denoising low-dose computed tomography (CT) images. This framework explicitly considers the unique noise characteristics of CT scans, moving beyond generic image denoising models. ENCORE reformulates noise synthesis based on a more realistic noise distribution and extracts local noise power and correlation contexts. It also introduces a FlyingConv module that adaptively adjusts convolution weights for different image regions, improving both denoising quality and computational efficiency. Additionally, ENCORE allows for zero-shot conditional denoising by manipulating noise context maps at inference time, enabling dynamic control over output image texture. AI
RANK_REASON The cluster contains a research paper detailing a new framework for image denoising. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
