In this context, thresholding of images of particles
on a variable background arizes as a subproblem Ð in
this case, falling particles (aggregates) having the sky
as backdrop. The strategy in this paper has been to
use a technique which is simple in the sense that a global
thresholding algorithm is iterated over a subset of
current objects according to as few, simple and powerful
rules as possible. The global thresholding algorithm
is hence, just a building block in an iteration process
which avoids complexities such as interpolation and
window size selection which feature in many adaptive
thresholding techniques (Chow and Kaneko, 1972;
PeÂrez and Gonzalez, 1987; Tseng and Huang, 1993;
Yang et al., 1994). The domain of images for which
the thresholding technique is suitable, is, generally
speaking, those where the background, although variable,
is locally brighter than the objects to be identi®ed,
in this instance, particles or clusters of particles.
When designing the thresholding algorithm, special
care was taken to choose, as a building block, a global thresholding algorithm that was particularly suitable
for images of aggregates.