Abstract:
Problem of texture analysis and discrimination using wavelet transform has been
under consideration for the last three decades by several researchers from different
fields. Unlike Fourier transform, several bases are available in wavelet transform for
signal decomposition, but none of these has been designed by considering the actual
texture images to be discriminated. Therefore, a system is required for the design of
wavelet bases which caters the actual texture image to be discriminated. There are
several factors involved in wavelet design which can greatly influence the desired
results such as regularity, length of the wavelet, orthogonal or biorthogonal etc.
In this research work, we have analyzed different regions of Pakistan on the
basis of their textural properties. We propose a design of wavelet bases using genetic
optimization, which will provide excellent discrimination between the multiple
texture image s. Our objective function is based on maximization of distinguishability
measure involving the computation of finer details subject to some wavelet
constraints. In contrast to well known orthogonal wavelet families, we have used extra
degree of freedom in design of the wavelet function for best possible texture
discrimination. In genetic optimization process, design parameters of wavelet are
optimized according to the characteristics of texture images under defined set of
constraints. Classification results of optimized orthogonal and biorthogonal wavelet
were compared with the existing wavelet families, which show that the results
obtained are superior in terms of texture discrimination. The proposed system is
iv, capable of designing optimized wavelet by changing the input texture images for
different applications such as medical images, satellite images, document analysis and
industrial application etc.
