Research map / Image generation / Autoregressive and latent generation
Double JPEG Detection in Mixed JPEG Quality Factors using Deep Convolutional Neural Network
Jinseok Park, Donghyeon Cho, Wonhyuk Ahn, Heung-Kyu Lee
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| insert quantization table information at the last conv output of the Barni network (range -60 to 60) (ablation) | Barni network · Double JPEG block detection (114,043 test blocks) | accuracy: 84.46 -> 89.89 | Improved |
| insert quantization table information at the first FC output (ablation) | Barni network (range -60 to 60) · Double JPEG block detection (114,043 test blocks) | accuracy: 84.46 -> 89.95 | Improved |
| insert quantization table information at the second FC output (ablation) | Barni network (range -60 to 60) · Double JPEG block detection (114,043 test blocks) | accuracy: 84.46 -> 86.79 | Improved |
| insert quantization table information at all three locations (ablation) | Barni network (range -60 to 60) · Double JPEG block detection (114,043 test blocks) | accuracy: 84.46 -> 90.37 | Improved |
| increase convolutional layers from three to four (ablation) | Proposed network · Double JPEG block detection (114,043 test blocks) | accuracy: 90.37 -> 91.83 | Improved |
| increase convolutional layers from four to five or more (ablation) | Proposed network · Double JPEG block detection (114,043 test blocks) | accuracy: 91.83 -> 91.33 | No clear effect |
| widen histogram range beyond -60 to 60 in the Barni network (ablation) | Barni network · Double JPEG block detection (114,043 test blocks) | accuracy: 84.46 -> 82.30 | Hurt |
| use histogram features of DCT coefficients instead of RGB pixels (ablation) | VGG-16 / Barni · Double JPEG block detection (114,043 test blocks) | accuracy: 50.00 -> 83.47 | Improved |
| use deeper histogram network (Barni) instead of Wang's network | Histogram CNN · Double JPEG block detection (114,043 test blocks) | accuracy: 73.05 -> 83.47 | Improved |
| add quantization table input with optimization (proposed*) instead of Barni network | Barni network · Copy-move manipulation detection (2100 images) | F-measure on copy-move: 0.6323 -> 0.7595 | Improved |
| add quantization table input with optimization (proposed*) instead of Barni network | Barni network · Blur manipulation detection (2100 images) | F-measure on blurring: 0.6450 -> 0.7783 | Improved |
Closest papers by meaning
- Towards Flexible Blind JPEG Artifacts Removal (ICCV 2021 · similarity 0.86)
- JPEG Artifacts Reduction via Deep Convolutional Sparse Coding (ICCV 2019 · similarity 0.84)
- Learning Dual Priors for JPEG Compression Artifacts Removal (ICCV 2021 · similarity 0.83)
- JPEG Inspired Deep Learning (ICLR 2025 · similarity 0.83)
- A New Benchmark and Model for Challenging Image Manipulation Detection (AAAI 2024 · similarity 0.80)
- JDEC: JPEG Decoding via Enhanced Continuous Cosine Coefficients (CVPR 2024 · similarity 0.80)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
