Detection of Doctored Images

Abstract

In this paper we address the problem of detection of image doctoring. Doctoring is a process of altering or modifying the contents of an authentic image with varied motivations. We propose two frameworks to address the problem of de- tection of image doctoring: (i) bi-spectral analysis (ii) cor- relation pattern of the Point Spread Function (PSF) using it- erative blind deconvolution (IBD). The most common types of the doctoring are splicing, cloning and re-touching. The detection of doctoring finds application in the area of digital forensic, law enforcement, medicine, security, military etc. In the first framework, using bispectral analysis, we capture the higher-order spatial content introduced due to the doctoring to detect the doctoring process. We use the deviation in the magnitude and the phase plots of bi-coherence to detect doc- toring. In the second framework we propose to use the cor- relation pattern of the estimated PSF in the detection process. In this case, we model the combined process of image cap- ture and doctoring as a linear filtering process. We use IBD to estimate the PSF of the filtering system. We demonstrate the proposed algorithms on different doctored images, which includes doctoring using splicing, cloning and re-touched im- ages.