一个低成本的二维双正交离散小波之超大型积体电路架构设计.PDFVIP

一個低成本的二維雙正交離散小波之超大型積 體電路架構設計 1 2 游 竹 陳少傑 1. 國立宜蘭技術學院電子工程系助理教授 2. 國立台灣大學電機工程系教授 摘 要 類似於短時間富立葉轉換，小波轉換提供了另ㄧ種信號處理的方法，特別的是它適合於空間及頻譜聚集性的分析。在本論文中，我們提出ㄧ個 低成本的二維雙正交離散小波轉換之超大型積體電路架構設計，藉由有效的安排計算排程及運用雙正交濾波器的對稱特性，因而得到硬體成本的最 2 小化。針對計算ㄧ個含L 長度的N ´ N 二維影像而言，我們所提出的架構需花費近N 時脈週期，另外要求約2NL - 2N 的存儲單元、如為偶長度的濾 波器則需15L/8個乘法器、如為奇長度的濾波器則需15ëL/2+1û /4個乘法器，以及4(L -1)個加法器等計算單元硬體。 關鍵詞 ：離散小波轉換、超大型積體電路 Design of a Low-Cost VLSI Architecture for 2-D Biorthogonal Discrete Wavelet Transforms Chu Yu1 and Sao-Jie Chen2 1. Assistant Professor, Department of Electronic Engineering, National Ilan Institute of Technology 2. Professor, Department of Electrical Engineering, National Taiwan University Abstract The wavelet transform, similar to the Short-Time Fourier Transform, furnishes an alternative approach to signal processing, especially suitable for the analysis of spatial and spectral locality. However, the 2-D wavelet transform needs the same enormous computation as their counterpart Fourier transforms. In this paper, we present a low-cost VLSI architecture for 2-D biorthogonal discrete wavelet transforms. Both the efficient arrangement of a computation schedule and the symmetric property of biorthogonal filters used in this architecture contribute to the hardware cost minimization. For the computation of an N  N 2-D image with a filter length L , this architecture spends near N 2 clock cycles, and requires about 2NL - 2N storage unit, 15L/8 multipliers for even-length filters or 15L/2+1 /4 multipliers for odd-length filters,