It is argued that accurate optical flow can only be determined if prob
lems such as local motion ambiguity, motion segmentation, and occlusio
n detection are simultaneously addressed. To meet this requirement, a
new multiresolution region-growing algorithm is proposed. This algorit
hm consists of a region-growing process which is able to segment the f
low field in an image into homogeneous regions which are consistent wi
th a linear affine flow model. To ensure stability and robustness in t
he presence of noise, this region-growing process is implemented withi
n the hierarchical framework of a spatial lowpass pyramid. The results
of applying this algorithm to both natural and synthetic image sequen
ces are presented. (C) 1994 Academic Press, Inc.