Application Of Embedded Zerotree Wavelet To The Compression Of Infrared Spectral

The aim of this thesis is to make infrared spectra database. Studies on the compression of infrared spectral data by wavelet transform are presented in this thesis. So far, as the part of the whole work has been finished as follows:First, the evaluated method of compressibility has been presented when applying wavelet compression to IR spectra. The method of controlling reconstructed RMSD can obtain the compression ratio under the condition of given reconstructed RMSD. Then the influence of IR spectral characteristics themselves on compression ratios has been profoundly studied. Finally the choice of wavelet functions that hold best compressibility has been researched. Four different wavelet families such as Daubechies wavelets, Coiflets wavelets, Symlets wavelets and biorthogonal wavelets at different order are selected to investigate. The results indicated that biorthogonal wavelets hold the best general compressibility.Embedded Zerotree Wavelet (EZW) algorithm was first applied to compress IR data in our work. The coefficients obtained by wavelet transform have two features. One is that the signal energies become concentrative. The other is that these wavelet coefficients have self-similarity. The famous EZW takes advantage of these two features so as to compress data. It was first proposed to deal with two-dimentional data such as images. In this thesis it was applied to compress one-dimentional data. We experimented with IR spectra of some compounds and furthermore made a comparison with another compression technique. We can expect that this technique is much better than others in terms of efficiently coding wavelet coefficients because the zerotree quantization is an effective way of exploiting the self-similarities of wavelet coefficients at various resolutions. The results showed that this is indeed the case.