Detection of Doctored Images using Demosaicing

Abstract

The advanced development in image editing software’s have made easy to doctor the digital images and they leave no visual clues. Digital doctoring has become a serious problem in various fields such as scientific publications, politics, internet etc. This has led to reducing trustworthiness and creating false beliefs in real world applications, hence it is necessary to protect against image doctoring using appropriate methods. The doctoring operations are performed using splicing, cloning and retouching. Digital image doctoring process introduces un-natural higher order correlations in doctored images. We detect these correlations using demosaicing artifacts. The proposed features are based on the differences in the acquisition process of images. More specifically variant of color filter array. The basic idea is that the correlation between bit planes of an image will be different in doctored image compared to real image.