Multi-scale LPQ-DCT for Image Forgery Detection

Nowadays, image processing tools with advanced and easy-to-use features are becoming available making image manipulation much simpler than ever before. With the wide spread use of images, image authenticity is crucial and it is essential to be able to passively detect forgery when the original image is not accessible. In this work, we proposed an image forgery detection model based on the variations of the discrete cosine transform coefficients of transformed images using local phase quantization at different scales. Texture features are extracted by the proposed method and employed to train a support vector machine for image classification. Variants of the proposed model are evaluated on two publicly available benchmark datasets for color images and compared to other existing methods. The results demonstrate that significant improvement can be achieved.