Identifying computer generated graphics VIA histogram features

Discriminating computer generated graphics from photographic images is a challenging problem of digital forensics. An important approach to this issue is to explore usual image statistics. In this way, when the statistical distributions (i.e., histograms) of some types of residual images are established, previous works usually apply operations on these histograms or compute statistical quantities to extract features. However, as the histograms are fundamental resources and can present most image information, the histograms themselves can be directly used as features and we do not need further manipulations on them. Based on this consideration, we simply take several highest histogram bins of the difference images as features to carry out classification, and these simple histogram features work well in terms of both detection accuracy and computational complexity. Actually, experimental results demonstrate that, with only 112 features, the proposed method outperforms some state-of-the-art works.