A new dataset of real JPEG quantization tables (DocQT) improves forgery localization robustness on benchmarks and operational documents when models explicitly condition on the quantization table.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
DocQT: Improving Document Forgery Localization Robustness via Diverse JPEG Quantization Tables
A new dataset of real JPEG quantization tables (DocQT) improves forgery localization robustness on benchmarks and operational documents when models explicitly condition on the quantization table.