A study of an efficient embedding algorithm for low-density generator matrix embedding

• Main Authors: Kai-Ren Chen, 陳楷壬
• Other Authors: Chi-Yuan Lin
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/33405377748998150451

碩士 === 國立勤益科技大學 === 資訊工程系 === 102 === This study investigated the belief propagation(Bias Propagation, BiP) algorithm proposed by Fridrich. Low-density parity-check [9] (LDPC) coding is a matrix embedding method. If a parity-check matrix with LDPC codes serves as a generator matrix and the BiP [1] algorithm is derived from this coding framework, the embedding efficiency and the computation speed of the BiP algorithm are superior to that of conventional matrix algorithms. In this study, MATLAB was used to simulate experimental data obtained from various algorithms. The results indicated that the embedding efficiency of the BiP algorithm surpassed that of other previously investigated algorithms. For the BiP algorithm, four parameters (i.e., gamma, iteration, damping, and generator matrix G) most easily affect the embedding efficiency of the algorithm. Furthermore, the method proposed by Richardson was used to design a new low-density generator matrix [2] (LDGM). Overall, simulations were performed to suboptimal parametric values, and possible combinations were simulated using LDGM with various embedding rates. The goal was to identify the optimal parametric values under various experimental data. In addition, regarding the BiP algorithm, a damping method that involves using arithmetic averages was proposed to reduce the computation complexity than that of the original damping procedure.