Research On Parameter Estimation Of Image Operator Chain Based On The Correlation Of Operations

With the rapid development of Internet technology and the popularization of various digital media devices,digital image has become one of the most mainstream information acquisition and transmission media in people's daily life.However,with the emergence of powerful and easy-to-use image editing software,many security problems related to image authenticity have become increasingly serious.Passive forensic technology of digital image is a kind of computer technology which can verify the integrity,authenticity and originality of digital image.It has gradually become a research focus in the field of information security.In this research field,the operation parameter estimation technology of digital image is helpful to understand the process of image tampering,which is very useful to approximate restoration of image.However,most of the existing parameter estimation techniques are designed for single operation,and the detection result is not ideal in the complex background that images were tampered by multiple operations.Therefore,this paper focuses the parameter estimation of image operator chain.The main research results are as follows:(1)A new framework based on operation order and forensics feature is designed to discriminate the correlation between operations in operator chains.The framework mainly includes the correlation discrimination between operations and the corresponding forensic scheme design.Based on the influence of operation topologies and parameters on the traces left by tampering operations,we analyze the histograms and forensic features of the final generated images.The correlation degree of each operation in an operator chain is quantitatively described by using corresponding theories,and then the correlation degree is divided into coupled and uncoupled.Based on the framework,according to the different combination of tampering operations,the corresponding parameter estimation schemes are adopted to improve the detection performance under different correlation degrees.(2)Several parameter estimation features are designed to detect specific operator chains.We analyze the correlation degree between contrast enhancement,median filtering and resizing,and then the corresponding forensic features are proposed based on different correlation degrees.For uncoupled operations: contrast enhancement and median filtering,we use the histogram feature of image spatial domain and the energy density feature of image frequency domain to estimate the operation parameters without considering the operation sequences.For coupled operations: median filtering and scaling,we adopt difference-image to highlight image details,enhance image frequency domain features,and expand the feature dimensions to improve the effect of supervised learning.The experimental results show that the parameters of the above operation chains can be estimated effectively by using the forensics methods and features proposed in this paper.