A deformable block-matching algorithm for tracking epithelial cells

摘要

A deformable block-matching algorithm based on the notion of a deformation tensor was developed to track embryonic cells from one time lapse-image to another and from one viewing angle to another. The algorithm makes use of a bi-linear transformation to geometrically deform the pattern between the source and target images and a cross-correlation, block-matching algorithm to select the best match. When used to track natural variations in surface pigmentation that occur on axolotl (Ambystoma mexicanum) embryos, depending on the image magnification, the algorithm will accurately track details of cell boundaries, or groups of many cells.