Superpixel-Based Watermarking Scheme for Image Authentication and Recovery
Based on the superpixel segmentation mechanism, a novel fragile watermarking scheme for image authentication and recovery is proposed. For each superpixel region, the authentication watermark is generated by putting the pixels into a feedback-based chaotic system and embedded into the region itself. The content of each superpixel is compressed to construct the recovery watermark, which is embedded into another selected superpixel. To extract the authentication and recovery watermark properly, the superpixel boundaries are marked. The reliability of a superpixel is first determined by its authentication watermark. To improve detection accuracy, the recovery watermark extracted from authentic superpixels is utilized for precise detection. Moreover, the recovery information extracted from authentic superpixels is decompressed to recover the tampered regions. Experimental results demonstrate that the proposed method can not only resist general counterfeiting attacks, especially vector quantization (VQ) attack, but also has an excellent performance on location accuracy and self-recovery.
KeywordsSuperpixels Chaotic system Precise detection Authentication Self-recovery VQ attack Watermarking
This work is supported in part by 973 Program (2011CB302204), National Natural Science Funds for Distinguished Young Scholar (61025013), National NSF of China (61332012, 61272355), PCSIRT (IRT 201206), and Open Projects Program of National Laboratory of Pattern Recognition (201306309).
- 2.Wong, P.W.: A public key watermark for image verication and authentication. In: Proceedings of the IEEE International Conference on Image Processing, Chicago, IL, pp. 425–429, October 1998Google Scholar
- 5.Suthaharan, S., Kim, S.: A gradient image dependent fragile watermarking for improved security and localization. In: Proceedings of the IEEE Southeast Con, pp. 343–348 (2004)Google Scholar
- 10.Ren, X., Malik, J.: Learning a classifacation model for segmentation. In: ICCV 2003, pp. 10–17. IEEE Computer Society (2003)Google Scholar
- 11.Fulkerson, B., Vedali, A., Soatto, S.: Class segmentation and object localization with superpixel neighborhoods. In: ICCV 2009, pp. 670–677. IEEE (2009)Google Scholar