Po-Yueh Chen, En-Chi Liao and Chung-Wei Liang
Department of Computer Science and Information Engineering
National Changhua University of Education
Changhua, 500 Taiwan
Discrete wavelet transform (DWT) is an efficient tool for multi-resolution decomposition of images. It has been shown to be very promising due to its high compression ratio and self-similar data structure. Conventionally a 2-D DWT is accomplished by performing two 1-D operations: one along the rows and the other along the columns of an image. Without executing ordered 1-D transforms, we develop a new algorithm to compute a 2-D Haar DWT, the simplest DWT. Two merits of this algorithm are compactness and quickness. The algorithm is implemented with a compact, regular VLSI architecture whose system throughput can be conveniently improved by appropriate parallel/pipeline methods.
Keywords: segmented-matrix algorithm, Haar discrete wavelet transform, VLSI architectures, parallel, pipeline
Retrieve PDF document (200807_18.pdf)
Received July 17, 2006; revised January 19 & March 20, 2007; accepted May 23, 2007.
Communicated by Ja-Ling Wu.