DOI: 10.14704/NQ.2022.20.15.NQ88004

High performance RNS based DWPT for Image processing application

D. Vijaya Sree , Dr. D.R.V.A Sharath Kumar

Abstract

The lifting-based wavelet computing paradigm emerged as a key alternative for trading off accuracy and energy efficiency. The lifting based Discrete Wavelet Transform (DWT) pipelined VLSI architecture with an arbitrary wavelet tree is used. Here memory utilization rate is optimally reduced using bypass and reordering models. The lifting discrete wavelet transform (LDWT) is a low-complexity pre-processing filter suitable for image processing systems which are incredibly energy constrained. This work proposes the replacement of conventional arithmetic with improved RNS arithmetic model in a DWT hardware architecture and also to implement this by using Image processing application. The use of this technique helps to improve the system performance in terms of path delay and energy efficiency which validates the performance metrics using image quality assessment (IQA) measurements. In summary: the results prove the energy and delay reduction with approximation and the highest acceptable compression ratio processing systems which are incredibly energy constrained

Keywords

approximate computing, discrete wavelet transform (DWT), VLSI design, RNS system, image processing.

Full Text

References