An analogy between colour and spatial coding.

Abstract

The early stages of colour coding are well established in that the trichromatic receptor stage is followed by a set of opponent colour channels. One interpretation of the sequence is that opponent channels carry unrelated aspects of the colour stimulus, unlike the cone channels. The overlap of the cone channels can be removed by decorrelating their spectral-sensitivity functions, and this procedure has been found to give opponent colour channels which match those found psychophysically. Since the known spatial-frequency channels also show considerable overlap, the question arises which aspects of the spatial stimulus are captured by decorrelating the spatial-frequency channels. The results of decorrelating the spatial-frequency channels are that the first decorrelated spatial filter acts as a broad bandpass filter which has a peak sensitivity at 7.9 cycles deg-1, and that the second decorrelated spatial filter acts as an opponent spatial-frequency channel, with a minimum output at a low (4.1 cycles deg-1) spatial frequency and a maximum output at a high (15.1 cycles deg-1) spatial frequency. The characteristics of the first decorrelated filter closely resemble the properties of the foveal perceptive field which have been used to explain the Hermann grid illusion. Thus, the decorrelation analysis produces a model for the functional organisation of the channel implementation at the neural and psychophysical levels, but which directly relates to the subjective appearance of the visual stimuli.