Colour-sensitive edge detection using hypercomplex filters

Colour edge detection procedures usually aim to find significant discontinuities between all adjacent homogeneous image regions, no matter what the colour. In this paper we present a colour edge detection procedure designed to find edges between homogeneous regions of particular colours. Initially the three-dimensional RGB colour space is modelled as the imaginary subspace of a four-dimensional, hypercomplex space. A hypercomplex filter is then used to highlight edges between a region of colour C1 to a region of colour C2. A 4-D, thresholding criterion is derived to identify the image pixels belonging to these edges. We demonstrate that this procedure can successfully find particular coloured edges in both synthetic and natural images.