We present two contour-based techniques, for computerized object recognitio
n that avoid the difficulties due to translations, rotations and scaling. O
ur techniques do not require any shape representation. The first technique
uses the scale-space filtered coordinate functions of contours and their "l
argest diameters". The second, uses the Hotelling transform of the vector r
epresentations of the points of contours. We applied both techniques to rec
ognize human corneal endothelial cells, embedded in a sample of tissue. The
results obtained using the Hotelling transform represent a considerable im
provement when compared to those previously obtained using the coordinate f
unctions, the curvature function and Fourier descriptors as representations
of shape. (C) 2001 Elsevier Science Ltd. All rights reserved.