A Novel Approach for Detecting Image Forgery

To identify the traces of various forensic problems is an important issue in digital image forensic. The method propose a novel approach for detecting the traces of JPEG compression and image tampering using statistical feature extraction method. Discrete Cosine Transform Residual Features are used for extraction process. The method includes JPEG compression and that provides quantization noise based solution. Multiple-cycle JPEG compression is performed for noise analysis and define a quantity called forward quantization noise. The method analytically derive that decompressed image have lower variance of forward quantization noise. Using 64 kernels of DCT the quantized feature sets are generated and is so called as undecimated DCT. The proposed method solves the problems such as revealing the traces of JPEG compression history and identifies the tamped images using simple yet very effective detection algorithm. For chroma sub sampling and for small images image size the method is robust. The proposed algorithm can be applied in many practical applications, such as Internet image classification and forgery detection.