Image Segmentation Using Matrix-Variate Lindley Distributions

The aim of this article is to study a statistical model obtained by the mixture of the Wishart probability distribution on symmetric matrices. We call it "Matrix-variate Lindley distributions". We show that this distribution includes the matrix-variate Lindley distribution of the first kind and second kinds on the modern framework of symmetric cones. Its statistical properties and its relationship with the Wishart distribution is discussed. For estimating its parameters, an iterative hybrid Expectation-Maximization Fisher-Scoring (EM-FS) algorithm is created. Finally, in medical image segmentation the effectiveness and the applicability of the proposed distributions are proved with respect to the Wishart distribution.