New zero-shot method localizes image manipulations using artifact patterns

Researchers have developed a novel zero-shot pipeline for localizing manipulated regions in color images. This method bypasses the need for training data or device enrollment by estimating a reference artifact pattern directly from the noise residual of a single image. The pipeline includes a denoiser selection criterion, a block-level correlation analysis, and a Gaussian Mixture Model scoring stage to generate a pixel-level tampering probability map. Comparisons against existing state-of-the-art passive methods indicate the proposed approach is competitive. AI

IMPACT This research could enhance digital forensics by providing a more robust and accessible method for detecting image tampering.

RANK_REASON The item is an academic paper detailing a new technical method. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• alphaXiv
• arXiv
• Bayer
• CatalyzeX
• DagsHub
• Edgar González Fernández
• Gaussian mixture model
• Gotit.pub
• Hugging Face
• ScienceCast

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →