This paper presents a hierarchical modular neural network for colour c
lassification in graphic arts, capable of distinguishing among very Si
milar colour classes. The network performs analysis in a rough to fine
fashion, and is able to achieve a high average classification speed a
nd a low classification error. In the rough stage of the analysis, clu
sters of highly overlapping colour classes are detected Discrimination
between such colour classes is performed in the next stage by using a
dditional colour information from the surroundings of the pixel being
classified. Committees of networks make decisions in the next stage. O
utputs of members of the committees are adaptively fused through the B
ADD defuzzification strategy or the discrete Choquet fuzzy integral. T
he structure of the network is automatically established during the tr
aining process. Experimental investigations show the capability of the
network to distinguish among very similar colour classes that can occ
ur in multicoloured printed pictures. The classification accuracy obta
ined is sufficient for the network to be used for inspecting the quali
ty of multicoloured prints.