Automated identification of copy-move forgery using Hessian and patch feature extraction techniques

Area of Science:

• Digital Image Forensics
• Computer Vision
• Image Processing

Background:

• Copy-move forgery is a common image manipulation technique where parts of an image are duplicated and reinserted.
• Detecting such forgeries is crucial for maintaining the integrity of digital evidence and media.
• Existing methods often struggle with subtle manipulations or post-processing alterations.

Purpose of the Study:

• To develop a robust method for detecting copy-move forgeries in digital images.
• To combine Hessian and Raw patch features for improved localization of cloned image regions.
• To evaluate the proposed method's effectiveness on standard datasets and under various image attack conditions.

Main Methods:

• A hybrid feature extraction approach combining Hessian and Raw patch features on gray-level images.
• Utilizing a model based on key points detected by the Hessian detector for feature extraction.
• Employing the parallelism condition and random sample consensus (RANSAC) for accurate matching and mismatch elimination.

Main Results:

• Achieved a 100% F1 score on the GRIP database for copy-move forgery detection.
• Obtained a 92.13% F1 score on the image manipulation dataset (IMD).
• Demonstrated high F1 scores even on images subjected to noise, JPEG compression, and scaling attacks.

Conclusions:

• The proposed hybrid feature-based method is highly effective for detecting copy-move image forgeries.
• The technique shows robustness against common image processing operations used to obscure forgeries.
• This approach offers a reliable solution for digital image forensics and integrity verification.