Rst-Invariant Digital Watermarking to Face Image Database

SH Liu,HX Yao,W Gao
DOI: https://doi.org/10.1142/9789812702654_0040
2004-01-01
Abstract:Based on Radial Template (RT) and spread spectrum technology, this paper presents a novel digital face image watermarking algorithm that is invariant to rotation, scaling, and translation (RST). Robustness is one of the crucial important issues in watermarking. Robustness against geometric distortion with blind extraction remains especially challenging. This paper utilizes human face images' characteristics, and then proposes a model named RT to locate rotated faces in any orientation, This template is designed to find stable features of center-rotated objects in edge maps. Based on skin detection and edge extraction, our method searches for face-like areas and get their orientations by RT searching. Then the candidates are rotated upright and a spread spectrum watermarking algorithm is used to embed watermark in face images. And this watermark system will be used for main applications: copyright of face image database, identification card's watermark, and portrait-right protection. The experiments demonstrate that the scheme is invariant to rotation and translation except scaling.
What problem does this paper attempt to address?