Improved SIFT Feature-Based Watermarking Method for IHC Ver. 5

The current scale-invariant feature transform (SIFT) feature-based watermarking method proposed by Kawamura and Uchida was evaluated using Information Hiding Criteria (IHC) ver. 5. The marked regions were selected around the SIFT feature points then normalized to a uniform size against scaling attack, and watermarks were embedded into these normalized regions. There are two problems with this method. One is that the marked regions are distorted by normalization. The other is that the feature points are not detectable due to embedding on the points. Therefore, we propose an improved version of this method to solve these problems. We introduce two improvements called as a gradual magnification factor for normalization and concentric square regions for marked regions. The proposed method was evaluated using IHC ver. 5, and the feature detection rate improved and the bit error rate decreased.