Model-based hierarchical clustering with Bregman divergences and Fishers mixture model: application to depth image analysis

摘要: Model-based clustering is a method that clusters data with an assumption of statistical model structure. In this paper, we propose novel model-based hierarchical for finite mixture based on the Fisher distribution. The main foci proposed are: (a) provide efficient solution to estimate parameters (FMM); (b) generate hierarchy FMMs and (c) select optimal model. To aim, develop Bregman soft FMM. Our estimation strategy exploits divergence agglomerative clustering. Whereas, our selection comprises parsimony-based approach evaluation graph-based approach. We empirically validate by applying it simulated data. Next, apply real perform depth image analysis. demonstrate can be used as potential tool unsupervised