A Study of Efficient and Controllable Data Hiding Algorithms for Point-Sampled Geometry

• Main Authors: Ching-Kai Jan, 詹清凱
• Other Authors: Chung-Ming Wang
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/08655054374898629049

碩士 === 國立中興大學 === 資訊科學與工程學系 === 96 === A common drawback in the literature of data hiding for 3D point-sampled geometry is that most algorithms are not able to precisely control the model distortion caused by embedding the secret message. This drawback may result in eavesdroppers’ suspicion, failing to achieve the goal of covert communication. In addition, most data hiding algorithms provide low embedding capacity. This thesis proposes two data hiding algorithms in spatial domain for point-sampled geometry. The first algorithm aims at achieving a compromise between controllable model distortion and the embedding capacity. The second algorithm intends to increase the data embedding capacity without enlarging the current magnitude of the model distortion. The first algorithm we propose is a controllable data hiding algorithm, where the model distortion and the embedding capacity can reach a compromise. First, users are requested to assign two parameters including Allowable Maximum Distortion Ratio (AMDR) and the Minimum Embedding Bit Per Interval (MBPI) before the data embedding. The parameter AMDR is used to restrict possible distortion caused by data embedding. The parameter MBPI represents the minimal hiding capacity that users intend to achieve. Second, given two parameters, our algorithm automatically adjusts them to appropriate values by analyzing the geometric features of the point-sampled geometry. Finally, the algorithm embeds the secret message based on the adjusted values. As a result, the stego model being produced comes to a compromise between the distortion restriction imposed by users and the maximal embedding capacity requested by users. The second algorithm we propose is an efficient data hiding algorithm for 3D point-sampled geometry. This algorithm is inspired from an efficient data hiding approach conducted for image steganography. We modify the approach and extend it to 3D models. Experimental results show that our algorithm can increase up to 12.5% of the embedding capacity yet producing the same magnitude of the model distortion as our counterparts. In conclusion, we propose two data hiding algorithms for point-sampled geometry in spatial domain. To the best of our knowledge, our algorithms are original in terms of the capability for distortion control and the high embedding capacity. Our algorithms have the following characteristics: controllable, efficient, high embedding capacity, flexible, lower distortion, and better performance.