Ali O. Little bit devices of wavelet-transform

This thesis devoted to design little bit devices of wavelet transform with the help of CAD systems. During design process, a problem of word length choice arises. Limited word length calculations produce quantization noise so we tried to investigate the influence of quantization noise in one-dimensional and two-dimensional wavelet transform realizations on microcontrollers on the accuracy of signals and image restoration. The aim of this thesis is to realize DWT on little bit microprocessors (microcontrollers). In this work, noise models built for mirror filters (filter bank) and filters of lifting algorithm, taking in account the type and the characteristics of the wavelet functions, and the noise resources in the filters, which arise from the process of quantization. In this thesis, a method for choosing the right digit according to the noise level was designed. A program written in MatLab for transforming the coefficients of mirror filters to coefficients of lifting filters, taking in account the effects of quantization. Experiments and simulation on MatLab show the possibility of little bit microprocessor-based realization of DWT. Finally, the program of lifting algorithm of DWT on microcontroller family Microchip 18Fxxx (18F452) was simulated in MPLAB IDE v7.60. The obtained experimental results were close to the theoretical ones.