Image Forgery Detection Based on Gabor Wavelets and Local Phase Quantization

Image Forgery detection has become a hot area of research as a result of the increasing number of forged images circulating around in the Internet and other social media and due to the legal and social issues that they are creating. The key problem faced by the researchers is to categorize an image as forged or authentic and to localize the forgery. Several methods were proposed, but a proper method which can accurately detect image forgery is yet to be invented. Here, we propose a novel image forgery detection technique by taking texture information of the image as a distinguishing feature. The method relies on Gabor wavelets and Local Phase Quantization (LPQ) which can extract relevant texture features which are fed as input to a Support Vector Machine for classification. The results indicate an accuracy of over 99% on both CASIA v1 and the DVMM color dataset. It outperforms similar state-of-art methods in solving image forgery detection.