After presenting some elementary criteria that should be respected by
any automatic shape analysis technique, this paper focuses on the impo
rtance of the binary image encoding method. Most image analysers simpl
y use a raster image to represent a binary object. If, occasionally, a
vectorial description is available, it is merely chosen for its perfo
rmances in data compression. Data compression and shape analysis have
different goals and usual methods cannot satisfy both. The calypter is
a new descriptor vectorizing the shape as a set of maximal inscribed
discs. It is the most efficient means of accomplishing Euclidean mathe
matical morphology transformations and allows for further developments
in binary shape processing. A simple adaptive contour filtering techn
ique is presented. The calypter offers local and global perception of
shape characteristics. It permits complete automation of the morphomet
ric roundness charts used in many laboratories and also generates new
shape parameters. A case study of three sand populations is presented
to show the pertinence of a new 'equivalent roundness' parameter.