Content-Based Watermarking by Geometric Warping and Feature-Based Image Segmentation

In this work, we present a new content-based watermarking approach that uses geometric warping to embed watermarks with high robustness to strong lossy compression. The issue of hard decisions related to content-based watermarking is discussed and it is explained why hard decisions can involve bit errors in the watermark extraction process. This work contains a solution to prevent hard decisions increasing the watermark performance. Therefore, we introduce a new feature-based image segmentation process with high robustness to lossy compression. On the basis of the segmentation, a watermark approach is proposed. Further, a secret key can be used to prevent unauthorized access to the watermark. The watermark extraction process does not need the original image. Our analyses of the watermark approach confirm the expected high robustness to strong lossy compression.