Improving RLRN image splicing detection with the Use of PCA and kernel PCA

Area of Science:

• Computer Vision
• Digital Forensics
• Machine Learning

Background:

• Digital image forgery, particularly image splicing, is a growing concern due to advanced manipulation tools.
• Existing image splicing detection algorithms struggle with high dimensionality, feature redundancy, and computational time.
• Restoring trust in digital images necessitates robust and efficient detection methods.

Purpose of the Study:

• To improve the performance of the run length run number (RLRN) algorithm for image splicing detection.
• To address limitations of existing algorithms in handling high-dimensional features and reduce computational complexity.

Main Methods:

• Applied two dimension reduction techniques: Principal Component Analysis (PCA) and Kernel PCA.
• Utilized the enhanced RLRN algorithm integrated with PCA and Kernel PCA.
• Employed Support Vector Machine (SVM) for classifying authentic versus spliced images.

Main Results:

• Kernel PCA demonstrated superior performance as a nonlinear dimension reduction method.
• Significant improvements were observed across R, G, B, and Y color channels, as well as in gray-scale images.
• The integration of Kernel PCA with RLRN effectively handled feature dimensionality and redundancy.

Conclusions:

• Kernel PCA is an effective nonlinear dimension reduction technique for enhancing image splicing detection.
• The improved RLRN algorithm, particularly with Kernel PCA, offers a more efficient and accurate solution for digital image forensics.
• This approach contributes to restoring trust in digital images by providing a robust detection mechanism.