#48: A combined matching and tensor method to obtain high fidelity velocity fields from image sequences of the non-rigid motion of the growth of a plant root


H. S. Jiang, K. Palaniappan, and T. I. Baskin

IASTED Int. Conf. on Biomedical Engineering, pgs. 159--165, 2003

detection, segmentation, features, tracking, convex relaxation, plants, biomedical

PlainText, Bibtex, URL, Google Scholar

Abstract

This paper addresses the problem of estimating the non rigid growth motion of plant roots. Several assumptions and constraints about the motion estimation are made and a min-cost objective function for robust matching method is constructed. To expedite the matching method, a tensor-based method to initialize the searching parameters is provided. The tensor confidence field is used as a mask to indicate the sites where coherent motion occurred. Consistency between the forward matching and the backward matching is used as a confidence test. Other confidence testimonies include the sharpness of the peak/valley of the objective function in the robust estimation procedure.