A study on optimally constructed compactly supported orthogonal wavelet filters

Computer Science and Information Systems

article

Yongkai Fan¹  Qian Hu¹  Yun Pan¹  Chaosheng Huang³  Chao Chen¹  Kuan-Ching Li⁴  Weiguo Lin¹  Xingang Wu³  Yaxuan Li¹  Wenqian Shang¹ 

[1] State Key Laboratory of Media Convergence and Communication, Communication University of China;School of Computer and Cyber Sciences, Communication University of China;School of Vehicle and Mobility, Tsinghua University;Dept. of Computer Science and Information Engineering ,(CSIE), Providence University

关键词: compactly supported orthogonal wavelets;    the least-squares method;    Gauss-Newton methods;    LMF method;   

DOI  :  10.2298/CSIS210410052F

学科分类：土木及结构工程学

来源: Computer Science and Information Systems

【 摘 要 】

Compactly supported orthogonal wavelet filters are extensively applied to the analysis and description of abrupt signals in fields such as multimedia. Based on the application of an elementary method for compactly supported orthogonal wavelet filters and the construction of a system of nonlinear equations for filter coefficients, we design compactly supported orthogonal wavelet filters, in which both the scaling and wavelet functions have many vanishing moments, by approximately solving the system of nonlinear equations. However, when solving such a system about filter coefficients of compactly supported wavelets, the most widely used method, the Newton Iteration method, cannot converge to the solution if the selected initial value is not near the exact solution. For such, we propose optimization algorithms for the Gauss-Newton type method that expand the selection range of initial values. The proposed method is optimal and promising when compared to other works, by analyzing the experimental results obtained in terms of accuracy, iteration times, solution speed, and complexity.

【 授权许可】

CC BY-NC-ND