First step towards parameters estimation of image operator chain

Abstract

Many effective techniques have recently been proposed to estimate the parameters of specific tampering operations. Most of them consider the situation in which an image is tampered with by only a single operation. However, in reality, multiple operations are used to falsify an image. Because the tampering traces of previous operations may be weakened or eliminated by later operations, it is difficult for the prior algorithms, each of which was developed for a single operation, to detect all tampering operations. In this paper, we propose a new method for estimating the parameters of operations in different manipulation chains. A framework is presented to investigate the correlation between multiple operations, which divides the degree of correlation into uncoupled and coupled operations. Then, two cases of certain operator chains with different degrees of correlation are adopted to reveal the assessment framework. Under this framework, we design well-directed features to estimate the parameters for each operator chain. Finally, the experimental results demonstrate the effectiveness of the proposed framework.