A Blind-based Patchwork Watermarking for Motion Capture Data

• Main Authors: Chieh Tsai, 才傑
• Other Authors: none
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/96805518596778154361

碩士 === 義守大學 === 資訊工程學系碩士班 === 97 === With the fast progress of digital technologies and graphic accelerator, people have been able to make photorealistic 3D animations. Over the past few years, motion capture technology has played an important role in human motion description. However, acquiring an ideal motion capture data often needs considerable work and expensive cost. Since offenders are easy to duplicate and reuse motion capture data, seeking an effective copyright protection technique becomes an urgent topic in these days. This thesis proposes a novel blind watermarking method, based on the patchwork algorithm, via embedding the secret information to the frequency domain of motion capture data. According with the human visual system, a distortion metric is adopted to evaluate the importance of each joint for the use of deciding the embedding amplitude of the joint. The statistical feature of data distribution is investigated to provide the evidence of the watermark. Some empirical tests are given to attack the watermarked data as well to evaluate the robustness of the watermarking.