Summary: | 博士 === 國立高雄第一科技大學 === 工程科技研究所 === 98 === Digital watermarking techniques provide a scheme for the copyright protection of multimedia data. In recent few decades, most digital watermarking and steganographic research paid much attention to images, audio or video, which are raster (or regular) data, but little research addressed irregular vector graphics data specified geometrically, by vertices, polygonal lines or high-order curves. Vector graphic images, applicable for scalable vector graphics (SVG) format or geographical information systems (GIS), are popular on the WWW, and provide the advantage of permitting affine transformations without aliasing. The creation of cartoon images or the acquisition of GIS geometric data involves much work, so the copyright and ownership of vector graphics data should be protected in advance.
Common components in vector graphic images are polygonal lines or polylines. This dissertation proposes a normal multi-resolution representation of a polygonal line, and embeds a copyright notice or serial number in this representation. Previous studies on polyline watermarking have the non-transparent problems, including self-intersection of line segments. This dissertation proposes a perceptually transparent watermarking approach for polylines, and solves the self-intersection problem. The watermarking approach is also resistant to similarity transformation, traversal reordering, point insertion/deletion and random noise attacks.
This dissertation also develops a novel watermarking approach suitable for street-network vector maps that records the information about roads by points and line segments. The proposed method simplifies the map using the Douglas Peucker algorithm to obtain the feature points, and then subdivides the map into mesh segments. After map segmentation, the watermark is embedded in each segment. The proposed approach is superior to other vector map watermarking methods, since it resolve the synchronization problem from the alternation of vertex coordinates. Moreover, the detection step does not require the original map in either the segmentation preprocessing or the watermark detecting.
These two novel digital watermarking approaches respectively for two types of vector graphics: one is a polyline for contour description, and the other is the type of street-network vector maps for GIS applications. Experimental results show that the polyline watermarking algorithm based on normal multi-resolution representation is robust to many attacks In addition; the watermarking approach for street-network vector maps is blind, and robust to simplification, cropping and other attacks.
|