Illustration watermarking: an object-based approach for digital images

For most applications common watermarking techniques usually spread the data to embed over the entire media, since distributing the watermark information promises an improvement in regard to security aspects, data hiding capacity or rather robustness in terms of redundancy. Distribution is controlled by a a visual or psychoacoustical model that takes limitations of the Human Visual System (HVS) and syntactical information about the signal characteristics into account. Therefore in most cases syntactical and not semantical aspects determine embedding. In our paper we introduce an approach for object based annotation watermarking which respects semantical characteristics of digital images, referred to as model for illustration watermarking. By applying a user-assisted segmentation process regions, representing semantical objects within the image, are identified and prepared for embedding. Providing robustness to typical image processing operations like cropping, scaling, compression and rotation, the proposed technique is applicable for binding additional illustrative information to selected objects within the medium. Moreover we identify the requirements of object based watermarking in consideration of imperceptibility, as well as watermark payload, and present first experimental test results.