A Generalized Ghost Detection and Segmentation Method for Double-Joint Photographic Experts Group Compression.

Area of Science:

• Digital Image Forensics
• Computer Vision
• Image Processing

Background:

• Digital image manipulation is widespread due to accessible tools.
• Joint Photographic Experts Group (JPEG) format is prevalent, necessitating robust forensic methods.
• Existing double JPEG detection methods lack robustness against common manipulations like grid shifts and require automation.

Purpose of the Study:

• To develop an improved and automated digital image forensics algorithm for detecting double JPEG compression.
• To enhance the accuracy and robustness of forgery detection in JPEG images.

Main Methods:

• A modified low-pass filter distinguishes textured areas.
• Inconsistencies in quality factors create a 'ghost image'.
• A novel segmentation method identifies ghost borders, and Kolmogorov-Smirnov statistic assesses authenticity.

Main Results:

• A simple, efficient algorithm for double JPEG compression detection is proposed.
• Sub-visual quality factor differences are revealed.
• The segmentation method achieved over 92% specificity and 75% precision.

Conclusions:

• The proposed method effectively detects double JPEG compression and image forgeries.
• Performance metrics show the algorithm outperforms six state-of-the-art methods.
• The automated pipeline offers a significant advancement in digital image forensics.