Image Splice Detection based on Illumination Inconsistency Principle and Machine learning Algorithms for Forensic Applications

Image tampering is a technique by which the image and visual properties are altered for many reasons like for enjoyment, for creating false evidence, for publicity etc. This can be done by the application of hardware, software and editing tools. A performance comparison and a novel approach for verifying the image authenticity through the exploration of the illuminant based transformed spaces which is given by Illuminant Maps (IM) is discussed in this paper. A method is proposed for detecting the image splice. For acquiring statistical-based information two algorithms are applied namely Generalized Grayworld Estimates (GGE) and Inverse Intensity Chromaticity (IIC). The various features thus obtained through image descriptors like color, shape and texture features. The proposed method provides a solution for detecting the image tampering. Two different data sets DA1 and DA2 are considered for the experiments and are tested. The feature extraction algorithms are tested namely SVM, KNN, GLCM and proposed method using GLCM and SURF. The proposed work produced a better result of 83.50%, 82.55% and 84.50% in terms of accuracy, sensitivity and specificity respectively comparing to the classical methods of image forgery detection.