Hardware and Throughput Efficient Architecture for Lifting-based Scheme for DWT

Abstract

This paper exhibits an outline of Discrete Wavelet Transform (DWT) designs and techniques. Two principle strategies; conventional convolution technique and lifting strategy have been examined. The lifting-method gives numerous advantages over conventional convolution strategy. The convolution-method processes double the number of wanted coefficients and discards half of them toward the end. This outcomes in twofold the number of desired computations. The lifting-method gives integer-to-integer and in-place wavelet transform. In this paper, a hardware, speed and throughput efficient architecture for two-dimensional DWT has been proposed. The architecture has been improved by the application of enhancement procedures like row-column, parallelism, pipelining, unfolding and folding. The 2D-DWT is yielded by either utilizing distinct 1D-DWT models over and over again. Different DWT designs have been compared with the proposed architecture based on resources used, throughput and computational speed.