Likeness-Based dissimilarity measures for Gaussian Mixture Reduction and Data Fusion