A new algorithm is proposed for representing and characterising the contour
s of 2D objects, as well as detecting their corners. The novelty lies in th
e curvature estimation technique which is based on circular histograms of t
he contour chain code. Two histograms are calculated at the neighbourhood o
f every point, and instead of considering them of linear type and being com
pared by correlation, they are considered to be circular and compared by a
new method which defines a circular mean of variable origin. The main advan
tages of this method are its low computational overhead and the precision o
f the obtained curvature functions.