Font Size: a A A

Research On Digital Watermarking Of 3D Geometry

Posted on:2007-04-08Degree:MasterType:Thesis
Country:ChinaCandidate:Y XieFull Text:PDF
GTID:2178360182486388Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Current watermarking technology focuses on those media types like text, images, video and audio streams. Just a little attention has been paid for watermarking of 3D models from researchers. With the rapid development of virtual reality, cooperative design and Web3D, more and more CAD data and digital products are entering the World Wide Web, and watermarking for geometry-based 3D models is becoming a new hotspot.This dissertation analyzes the basic principle and methods of watermarking for 3D geometry models. Research on digital watermarking algorithms for the 3D mesh model in the spatial domain and 3D shape model defined as a point set is made. The main works we made is as follows:A new watermarking scheme for a 3D mesh model has been proposed. After proper positioning and alignment of the 3D mesh models, a low complexity and effective masking operation is used for modifying imperceptibly the location of vertices. Before extracting and detecting watermark signals, mesh alignment and registration has been implemented in order to improve the exactness of watermark extraction. Experimental results show the ability of the proposed method to geometric transformation, and usually can survive the cropping attack and can improve robustness against additive random noise and mesh simplification.Using the capacity of GCS to map features, a learning algorithm has been improved to adapt the application of feature recognition of a point set, which uses normal counter storing information about the variation of the normals of the neural mesh during the learning process. A vertex is designated as feature if its normal counter exceeds a threshold. A new watermarking scheme for a 3D shape model defined as a point set has been proposed. The algorithm is based on GCS and the statistical analysis of its learning process for a point set. The watermarks embedded by using the method can be detected, to some extent, after such attacks as cropping, simplification and additive random noise applied on the watermarked point set model.
Keywords/Search Tags:digital watermarking, 3D mesh, neural network, point-sampled geometry, affine transformation
PDF Full Text Request
Related items