Due to its heavy-tailed and fully parametric form, the multivariategeneralized Gaussian distribution (MGGD) has been receiving much attention formodeling extreme events in signal and image processing applications.Considering the estimation issue of the MGGD parameters, the main contributionof this paper is to prove that the maximum likelihood estimator (MLE) of thescatter matrix exists and is unique up to a scalar factor, for a given shapeparameter \beta\in(0,1). Moreover, an estimation algorithm based on aNewton-Raphson recursion is proposed for computing the MLE of MGGD parameters.Various experiments conducted on synthetic and real data are presented toillustrate the theoretical derivations in terms of number of iterations andnumber of samples for different values of the shape parameter. The mainconclusion of this work is that the parameters of MGGDs can be estimated usingthe maximum likelihood principle with good performance.
展开▼