Abstract
Given the pivotal role of digital images in our daily lives, it is important to detect copy-move forgery (CMF) of digital images. The detection of CMF is often based on feature detection and matching. For feature matching, the existing algorithms make use of the Euclidean distance. For feature detection, the Haar transform is one of the most popular techniques. This study retrieves image features through the Haar transform, and then simplifies the features by principal component analysis (PCA). After that, false boundaries were detected, localized, and removed. On this basis, the texture features of the input image were analyzed, using the gray-level co-occurrence matrix (GLCM). Finally, Euclidean distance was utilized to match features, and the mismatched features were labeled as forgeries. Then, the proposed approach was simulated in MATLAB, with accuracy as the performance metric. The simulation results show that our approach outperformed the PCA by 13.6% in accuracy.