A method of representing an object appearing in a still or video image, by
processing signals corresponding to the image, comprises deriving the
peak values in CSS space for the object outline and applying a non-linear
transformation to said peak values to arrive at a representation of the
outline.