Keypoint clustering using cortical colour opponency

Coloured regions can be segregated by using colour-opponency mechanisms, colour contrast, saturation or luminance. This paper focuses on clustering coloured regions by using colour tuned end-stopped cells. Colouinformation is coded in separate colour channels to convey differently coloured regions. Then, by using multi-scale cortical end-stopped cells tuned to colour, this information is coded in all channels by mapping it to the multi-scale peaks. Finally, unsupervised clustering is achieved by analysing the branches of these peaks and by linking them together on the basis of their colour, saturation and luminance information