Statistical modeling of texture wavelet coefficients

This paper focuses on texture analysis using the wavelet transform. The statistical modeling of the wavelet coefficients with the Generalized Gaussian Distribution (GGD) raises the problem of estimating its parameters. Our aim is to examine the estimated values which enable the fitting of the wavelet coefficients histograms to a probability density function (PDF) of the GGD. The experimental results show that in most situations the fitted PDF does not approach the special cases of this family of distributions. However, results obtained in some cases lead us to think to the Laplacian distribution as a suitable one.