IMAGE COPY-MOVE FORGERY DETECTION USING LOCAL FEATURES AND CLASSIFIER-BASED APPROACH

Abstract

ABSTRACT: In contemporary environments, digital photographs are a prevalent and effective communication instrument. They have a substantial impact on the communication and information technology sectors. In the contemporary technologically sophisticated society, the phrase "seeing is no longer believing" is a statement of fact. Therefore, forensic scientific testing poses a significant obstacle to detection. The development of methods to identify and categorize copy-move forgery images for use in forensic investigations has garnered increased attention in recent years. This work suggests a method for detection and classification that employs SIFT and SURF in that order, with Ant Colony Optimization in the matching phase and block-based features in the feature selection phase. We employ support vector machines for SIFT and SURF features, as well as the proposed SIFT with ACO features, in addition to classification using Gaussian and poisson kernels. The results indicate that SIFT with ACO outperforms other methods.