Spatio-temporal life cycle of meteorological structures is an important par
t of study of routine numerical weather analysis. In this paper we present
an algorithm to track tracer clouds in INSAT image sequence. Given a pair o
f sequence images taken at a 30 min interval, the objective of cloud tracki
ng is to derive the path of movement of a cloud contour in the source image
to the corresponding contour of the destination image. This has direct rel
evance to cloud motion vector (CMV) analysis by which wind speed and direct
ion are estimated. We have utilized an improved algorithm to generate a set
of initial estimates of possible cloud motion vectors which are then filte
red through a shape-based approach. The contour of cloud mass is modelled a
s a perfectly flexible string, and depending on initial estimates of CMV, t
he source cloud contours are deformed. The CMV direction for which the defo
rmed contour best matches the destination contour, gives the optimum path o
f cloud movement from source to destination image. The algorithm is demonst
rated using images from INSAT.