This paper introduces an algorithm for critical point detection in tex
tured fluid flow images. A new measure is defined, based on dynamical
system properties, that identifies candidate critical points in an ori
entation field. The candidates are verified or rejected based on estim
ates of the local flow field properties. The algorithm can locate part
ially occluded and degraded flow structures, and applications of this
algorithm to experimental flow imagery are included. The algorithm per
formance is quantified, and it is compared to other detectors. (C) 199
7 Pattern Recognition Society. Published by Elsevier Science Ltd.