Desarrollo de un nuevo modelo de visión del color basado en la fisiología del sistema visual

Resum

We propose a new colour vision model (ATTD) which computes the perceptual descriptors of a chromatic test from its XYZ tristimulus values and those of the background that surrounds it. We have suggested some correlations between the mechanisms of the model and the physiology of the visual system. ATTD has a trichromatic stage and three opponent stages at precortical, striate cortex and extrastriate cortex levels. The trichromatic stage responses are combined yielding an achromatic Magnocellular channel and three opponent Parvocellular channels: two asymmetric ones of the red-green type and one of the blue-yellow type. The three opponent channels generate at cortical level the achromatic, red-green and blue-yellow perceptual channels. The model has gain control mechanisms at the receptor stage and in the lateral geniculated nucleus (LGN), a subtractive mechanism in the striate cortex, and static non-linearities of the Naka-Rushton type. Numeric values of the parameters of the model have been determined by minimising a coefficient of agreement between a small set of experimental data and the predictions of the model for these data. We have tested the performance of ATTD by comparing its predictions to those of the ATD95 and Hunt models and with several sets of experimental data which were not included in the training process. The chromaticity of the perceptual white with equal energy background coincides with the equal energy white for a wide range of luminances. For high test luminances some anomalies appear in the loci of the perceptual descriptors. The rings of constant Munsell chroma are more circular and with more equidistant hues than those of the other two models. ATTD accounts for a wide variety of effects related to the dependence between perceptual descriptors and colorimetric parameters. The models predictions of the achromatic locus are consistent with the conclusions of Werner and Walraven. In summary, ATTD accounts for a wide range of colour appearance phenomena with a small set of training data; therefore, the potential of the model lies in its structure, based on the physiology of the visual system.