A linear transport-based embedding method for analyzing biomedical images

A fundamental difficulty facing those who wish to extract quantitative information from biomedical images is finding a manner with which to compare two morphological examplars (e.g. cells or sub cellular structures). We present a new approach for finding a linear embedding for a set of images that is isometric to a transportation-based metric. The approach is capable of quantifying both shapes as well as textures, is generative, and facilitates the use of geometric data processing techniques such as LDA and PCA.