Spatial-frequency-tuned mechanisms of the red-green channel estimated by oblique masking

The sustained spatial-frequency-tuned (SF-tuned) mechanisms of nonoriented units were examined by means of orthogonal masking for the Red-Green (R-G) color channel, and those of oriented units by oblique masking for the achromatic channel but not for the color channels. An oblique-masking technique minimizes the artifacts that are due to spatial phase effects, local cues, spatial beats, spatial probability summation, and changing criteria. Therefore the spatial characteristics of the R-G color channel are now investigated by an oblique-masking technique and linked with my paper on orthogonal masking [J. Opt. Soc. Am. A 15, 1 (1998)]. The R-G channel was defined by the minimum-flicker and hue-cancellation techniques. A color monitor system was used to generate spatially localized (D6) vertical color test patterns [0.063-8 cycles per degree (cpd)] and sinusoidal oblique color masks (0.031-16 cpd, 1.2-60% contrasts). Color contrast sensitivity functions (CSFs), threshold elevation (TE) versus mask SF (TvSF) curves, and TE versus mask contrast (TvC) curves were measured by the method of constant stimuli with a two-interval forced-choice technique by using Powell's achromatizing lens under sustained (Gaussian, 2-s-duration) conditions. Results show the following: (1) The color CSF is a low-pass function of SF with average half-height SF of 0.7 cpd and cutoff SF of 14 cpd with the use of a color-detection criterion. (2) TvSF curves are broadly bandpass and fall into five groups, peaking at approximately 0.13, 0.5, 2, 4, and 8 cpd. The root-mean-square cone-color CSF is 3.8-5.4 times the stimulus-color CSF. (3) A "crowding effect" similar to that of the TvSF curves of the achromatic channel was also found, but the TvSF curves of the R-G channel are not sharply peaked, similar to the result for orthogonal masking. Data analysis led to the following conclusions: (1) A simple multiple-mechanism model yields one low-pass color mechanism (with average half-height SF of 0.54 cpd) and five bandpass SF-tuned color mechanisms; these six mechanisms are necessary to explain the CSF, TvSF, and TvC data simultaneously. (2) The bandpass mechanisms peaked at approximately 0.13, 0.5, 2, 4, and 8 cpd with average full bandwidths at half-heights of 3.6, 3.2, 2.1, 1.2, and 1.3 octaves, respectively. (3) Since oblique-masking color mechanisms (unlike achromatic oriented mechanisms) have broad orientation tuning under sustained conditions and there is a significant orthogonal masking, the oblique-masking color mechanisms may have contributions from both oriented and nonoriented units. (4) The high degree of similarity between the SF-tuned filters of mechanisms derived from oblique- and orthogonal-masking data suggests that most of the chromatic SF tuning is already accomplished by nonoriented units. (5) The quality of the fit to oblique- and orthogonal-masking data combined dropped enough to reject the hypothesis that the former taps the performance of only the same nonoriented mechanisms as those by the latter. Adding gain parameters that reduce the TEs for orthogonal masking gave a better fit, suggesting that orientation gains are one of the factors involved in the transformation of information from nonoriented to oriented mechanisms. However, the fit was still worse than that for oblique-(6) Since masking-alone or orthogonal-masking-alone data, suggesting that more factors may be involved. primate parvo lateral geniculate nucleus (pLGN) units behave in a fairly linear manner, the color contrast nonlinearity (which follows the linear filter) of a mechanism may be post-pLGN.