Laplacian Mixture Model Point Based Registration (2006.12582v1)

Abstract: Point base registration is an important part in many machine VISIOn applications, medical diagnostics, agricultural studies etc. The goal of point set registration is to find correspondences between different data sets and estimate the appropriate transformation that can map one set to another. Here we introduce a novel method for matching of different data sets based on Laplacian distribution. We consider the alignment of two point sets as probability density estimation problem. By using maximum likelihood methods we try to fit the Laplacian mixture model (LMM) centroids (source point set) to the data point set.