Efficient architectures for high Speed and low power vlsi Implementation of lifting discrete Wavelet transform

Abstract

The Discrete Wavelet Transform DWT plays a major role in the newlinefields of signal analysis computer vision object recognition and image and newlinevideo compression standards The advantage of the DWT over other newlinetraditional transforms is that it performs the multi resolution analysis of newlinesignals with localization both in time and frequency The JPEG 2000 standard newlineadopts two methods to produce the wavelet transform The frequency domain newlinemethod uses convolution for implementing the filter banks and the lifting newlineschemes are based on the time spatial domain representation of the sub band newlinecoding of the given image coefficients newlineThe implementation of the DWT in real time image video newlineprocessing has some issues First the computational complexity of the newlinewavelet transform is several times higher and it has to process massive newlineamounts of data at high speeds Second the DWT needs extra memory for newlinestoring the intermediate computational results The use of the software newlineimplementation of the DWT provides flexibility for manipulation but it may newlinenot meet the timing constraints in certain applications The high cost of newlinemultipliers has practical limitations in the hardware implementation of the newlineDWT The Filter bank implementation of the DWT contains two FIR filters It newlinehas traditionally been implemented by convolution or the finite impulse newlineresponse FIR filter bank structures Such implementations require both a newlinelarge number of arithmetic computations and storage which are not desirable newlinefor either high speed or low power image video processing applications newline newline