Improved Geometric Warping-Based Watermarking

Abstract

In this paper, we propose a blind geometric warping-based watermarking for images and videos. We put emphasis on high robustness and security as well as decreased complexity. Our previous works have shown that the geometric structure of images and videos is a suitable embedding domain for watermark information. To describe the geometric structure, detection or segmentation methods are used, which can result in additional security holes, in more required computing effort, in loss of robustness and in loss of watermark blindness. In this work, we propose a watermarking approach that doesn¿t use detection and segmentation processes. Our watermark approach is based on the result of a filtering that is used to describe the geometric structure. Additionally, an efficient realisation of the filter is developed and a scheme of watermark embedding is proposed. The watermark bits are spreaded over the whole frame and coded in the spatial positions of edges and corners. We show that the resulting watermark can be embedded without visible quality degradations. Beside of this, it achieves high robustness to strong lossy compression and other distortions at a high watermark capacity.For example, a H.264/AVC compression that degrades the video quality down to PSNR of 20.6 dB induces a watermark Bit Error Rate of 0.0013 at a watermark capacity of 20 bits per frame.