Detection of interpolation using correlation coefficients

Resampling detection is a very useful tool in digital image forensic. In this paper, we propose a new algorithm for detecting interpolation by using the periodicity characteristics hidden among the rows(or columns) of the image. The core of our algorithm is the use of correlation coefficients based on the observation that the correlation coefficients of the row(or column) vectors vary periodically in resampled image. In some images, the image contents themselves correlate periodically. To reduce this kind of interference and enhance the detection efficiency, we use the first derivatives signals and detect the image both horizontally and vertically under some condition. As experiments show, the proposed algorithm works well, and most of the resampling factors can be correctly determined with high accuracy.

I. Introduction

Over the past years, image tampering has become a serious problem as we enter the information era. To provide information security, passive forensic technique proves to be very useful in the fields of digital image authentication as it doesn't need any prior operations such as watermarks embedding. Among all others, resampling detection is a very important subject since resampling operation is often involved when using image splicing to create a consistent image and cover the trace of forgery. Resampling normally contains an interpolation step, which is an algorithm to produce new values using existing values. This step inevitably brings periodicity into the new generated image signals. Digging out the statistic changes may uncover the forgery and signify tampering.