Design and Implementation of One-Dimensional and Multi-Dimensional Variable Digital Filters

• Main Authors: Yun-Da Huang, 黃韻達
• Other Authors: Soo-Chang Pei
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/09573240339759174946

博士 === 國立臺灣大學 === 電信工程學研究所 === 103 === Variable digital filters have received considerable attention in the last two decades due to their wide usages in communication systems and image processing. Their important advantage is that they contain a number of parameters in transfer function, which can be used to tune the phase response immediately on-line without redesigning a new filter. In this dissertation, several 1-D variable fractional-delay (VFD) digital filters are proposed, such as minimax design, new criterion FIR and IIR filter design. After that, two structures for the design of 2-D VFD digital filter are illustrated in detail. Then, the application of these VFD filter is introduced which can be used in the field of digital notch-like filter designs. In this dissertation, we proposed not only the design of digital 1-D and 2-D VFD and notch-like filters but also the design of variable 1-D and multi-dimentional digital filters such as 1-D maximally linear FIR differentiators, 2-D fan type filters, 2-D circular symmetry filters, 2-D elliptical filters and 3-D cone-shaped filters by using McClellan transformations and multi-dimensional quadrature mirror filters. In the design procedure, several kinds of approaches are used, such as weighted least-squares(WLS) design, minimax design and iterative design whicn can achieve high performances of frequency responses and insure the stability of the designed IIR filters. In all types of digital filter designs, the related transfer functions and structures are illustrated which are the latest methods for these filter designs and can obtain the better performances compared with the present methods. Furthermore, a new method for the 2-D image filtering is also introduced which can obtain higher signal-to-noise ratios (SNR) in the output than the conventional methods.