Wasserstein KMeans for Clustering Tomographic Projections
Abstract
Motivated by the 2D class averaging problem in singleparticle cryoelectron microscopy (cryoEM), we present a kmeans algorithm based on a rotationallyinvariant Wasserstein metric for images. Unlike existing methods that are based on Euclidean ($L_2$) distances, we prove that the Wasserstein metric better accommodates for the outofplane angular differences between different particle views. We demonstrate on a synthetic dataset that our method gives superior results compared to an $L_2$ baseline. Furthermore, there is little computational overhead, thanks to the use of a fast lineartime approximation to the Wasserstein1 metric, also known as the Earthmover's distance.
 Publication:

arXiv eprints
 Pub Date:
 October 2020
 arXiv:
 arXiv:2010.09989
 Bibcode:
 2020arXiv201009989R
 Keywords:

 Computer Science  Computer Vision and Pattern Recognition;
 Electrical Engineering and Systems Science  Image and Video Processing;
 62H30 (Primary) 92C55;
 68U10 (Secondary);
 I.5.3;
 I.4.0
 EPrint:
 11 pages, 5 figures, 1 table