Convolution-thresholding is a new approach to describing interface motion t
hat unifies and generalizes Huygens' principle, threshold growth cellular a
utomata, and reaction-diffusion equations. Convolution methods have many de
sirable properties, including automatic capture of topological change, prod
uction of curvature motion without explicit computation of curvature, natur
al extension to the motion of triple-point junctions, and fast, accurate im
plementation. In this paper, we summarize the relation of convolution-thres
holding schemes to previous methods, and we review the theoretical and algo
rithmic development of this approach. We also review recent applications to
computer vision, developmental biology, excitable media, and material scie
nce. (C) 2001 Academic Press.