Abstract
Introduction: What preattentive mechanisms are sensitive to scrambles (spatially random mixtures) of colors from the L-M cardinal axis of DKL space? Methods: On a trial, the participant strove to localize a circular target patch of color-scramble with one histogram in a large, annular background color-scramble with a different histogram. Scrambles were composed of eight motion equiluminant (ME) lights, varying from saturated red to gray to saturated green along the L-M axis of DKL space. For each of two orthogonal histogram perturbations lk, k=1,2, the participant was tested in two (separately blocked) complementary attention conditions: in the first attention condition, the target histogram was dominated by lk and the background was dominated by -lk; in the second, the target histogram was dominated by -lk, the background by lk. Results are modeled by assuming that (1) the participant has a fixed set of mechanisms, each of which is activated nonnegatively by different colors, and (2) in a given attention condition the participant uses an optimal linear combination (with weights constrained to sum to 1) of only those mechanisms more strongly activated by the target than by the background. Results: The data can be accounted for with two preattentive mechanisms: one that assigns oppositely signed values to reds than to greens, and one that assigns oppositely signed values to highly saturated colors than to desaturated colors. The participant can selectively attend either to the positive or the negative responses of either mechanism. Conclusions: Human vision possesses two preattentive mechanisms sensitive to variations along the L-M cardinal axis of DKL space, one sensitive to mean color and another sensitive to saturation variance.
Meeting abstract presented at VSS 2015