Color Contrast Detection in Spatial Chromatic Noise
In this report, higher order color mechanisms for the detection of spatial distributions of color are investigated. The objective of this work is to understand which are the visual mechanisms involved in chromatic detection tasks and to dene the properties of such mechanisms. Despite great investigation eorts, human color vision is far from being satisfactorily explained. It is still not clear which mechanisms, at a post-receptoral level, mediate color detection. Recently, psychophysical studies have argued that the cone-opponent signals are further processed by higher order chromatic mechanisms, sort of multiple channels tuned to a variety of directions in the color space. A better understanding of chromatic detection mechanisms could enable the development of perceptual image quality metrics and the evaluation of chromatic noise masking eects. The mechanisms responsible for the detection of monochromatic signals have been described developing a psychophysical experiment. Such mechanisms have been characterized estimating contrast thresholds for the detection of a chromatic signal within a color spatial distribution using a sectored noise masking technique. The experiment was designed in the DKL color space, thus both signal and noise mask were dened in that space. Sectored noise draws samples from a sector of variable width in an equiluminant plane of the DKL color space. Such a sector is oriented along the same chromatic axis of the signal. Sector amplitude and width and test color direction were under experimental control. Observers were asked to detect a signal, a monochromatic Gaussian pulse, within a sectored noise. Contrast thresholds for the detection of red, light reddish, orange and yellow signals were measured. Estimating the potency of noise masking as a function of noise sector width, it is possible to distinguish directly between detection mechanisms that combine the photoreceptoral inputs in a linear or nonlinear fashion. In fact, a mechanism tuned to a certain color direction that combines the cone signals linearly is not inuenced by the noise components orthogonal to such a direction, that is, by the width of the noise sector. On the other hand, nonlinear mechanisms should be aected by the variation of sector width. In this report we carefully detail all the steps followed to develop our psychophysical experience. In our experiment, thresholds were found not to depend on noise sector width, consistently with the hypothesis of linear mechanisms operating for the detection of specic colors.