Image splicing detection with dwt domain extended Markov features

Abstract

Recently, image forgery has increased with the widespread of using digital images. At least two different images are used when forged images are generated in image splicing which is an image forgery method. Forged and original images are detected with features extracted from the image. In this study, a Discrete Wavelet Transform and Markov based method is used to extract/construct feature vectors. Firstly, with applying Discrete Wavelet Transform to the image low frequency component image is obtained, later four-direction Markov features are extracted from the low frequency component image. After the feature extraction, images are classified as forged and original with support vector machines. The method has been tested for accuracy ratio in the commonly used data sets (Casia v1.0, Casia v2.0, Columbia DVMM) in Image Splicing Detection. In the experimental results, it was seen that forged and orjinal images are detected with high accuracy ratio compared with the similar methods in the literature.