Abstract—An image splicing can be easily made by a skilled image editor, and verification of authentic image has become important. In this paper, we propose a detection method for splicing boundaries that uses uniformity on camera response function (CRF) in an image. This uniformity is observed as differential features of CRF. In altered image, the non-uniformity of CRF mostly appears on the edges of a splicing region. The characteristics of CRF extracted from edges in an image differ between an altered region and an authentic region. We analyze these characteristics by using QR map. The QR map represents the relation between this features and image intensity. An authentic boundary has the characteristics of a common CRF within an image, but a splicing boundary is not so. We show the tendency of authentic boundaries in QR map, and we argue about the method of estimating whether a target boundary is spliced by evaluating the consistency of QR map.

Index Terms—camera response function, splicing boundary, QR map, digital forensics

Cite: Hirokazu Kato, Toshiyuki Nagao, Hideaki Fujiwara, and Katsuya Kondo, "Boundary Detection of Altered Region by Differential Features on Camera Response Function," International Journal of Signal Processing Systems, Vol. 2, No. 1, pp. 31-36, June 2014. doi: 10.12720/ijsps.2.1.31-36