Abstract
Aim
Traditionally, thresholds for discriminating colour and luminance differences have been measured using stimuli such as disks, gratings or gabors, and accounted for in terms of the responses of relatively low-level mechanisms in the visual pathway. On this basis we would not expect the higher-order structure of, for example, natural scenes to be a factor determining colour and luminance discrimination thresholds. We therefore decided to compare discrimination thresholds between natural scenes and phase-scrambled versions of the same scenes.
Method
The stimuli were fifty calibrated color photographs of everyday scenes and fifty phase-scrambled images. The chromaticity and saturation of every pixel was represented as a vector in a modified version of the MacLeod-Boynton color space, and could be translated, rotated, compressed, or randomly repositioned within that color space. Thresholds for detection of each type of transformation were measured using a two-alternate forced choice method.
Results
Thresholds for all types of transformations in color space were significantly lower in natural scenes compared to phase-scrambled images. Thresholds for detecting random changes in color, in the form of either Gaussian or fractal noise, were considerably lower in natural compared to phase-scrambled images.
Conclusion
The structure of natural scenes plays a significant role in our ability to discriminate colour and luminance differences.
Supported by Canada Institute of Health Research grant #MOP-11554 given to F.K.