Blind Image Forgery Detection Using LBP and Statistical Moments

At present time, the ubiquitous availability of digital devices has rendered the validity of image contents doubtful via numerous open source and commercial image altering applications. The most prevalent forgery techniques are copy-move forgery and image splicing. As a result, methods for authenticating image contents are necessary. Local Binary Patterns (LBP) and Discrete Wavelets Transform (DWT) are presented here to validate the images. The input image is first transformed into YCbCr channels. Then, chroma channels are utilized to extract features. By capturing 4 statistical moments, features are retrieved using a 2-level DWT. For training and testing, an ensemble classifier is utilized and further classifies the images in authentic or forged. The suggested model provides high detection accuracy with relatively low dimensionality and testing time which demonstrates its efficacy.