Saliency interactions between the ‘L-M’ and ‘S’ cardinal colour directions

Abstract

Two sub-systems characterize the early stages of human colour vision, the 'L-M' system that differences L and M cone signals and the 'S' system that differences S cone signals from the sum of L and M cone signals. How do they interact at suprathreshold contrast levels? To address this question we employed the method used by Kingdom et al. (2010) to study suprathreshold interactions between luminance and colour contrast. The stimulus employed in one condition was similar to that used by Regan and Mollon (1997) for studying the relative 'organizing power' of the two sub-systems, and consisted of obliquely-oriented red-cyan (to isolate the L-M sub-system) and violet-chartreuse (to isolate the S sub-system) stripes within a lattice of circles. In our experiment there were two conditions, (1) the Separated condition, in which the L-M and S modulations were of opposite orientation and presented separately as a forced-choice pair, and (2) the Combined condition, in which the L-M and S modulations were added. In the Separated condition the task was to indicate the stimulus with the more salient orientation structure, whereas in the Combined condition the task was to indicate the orientation that was more salient. Psychometric functions were used to estimate the ratio of L-M to S contrast at the 'balance-point' i.e. point-of-subjective-equality (PSE) in both conditions. We found that across 20 subjects an average of 8% more S than L-M contrast was needed to achieve a PSE in the Combined compared to Separated condition. We consider possible reasons for this PSE difference and conclude that it is either due to an early-stage interaction between the S and L-M sub-systems, or to a later stage in which new colours that arise from their combination are selectively grouped.

Texture Variations Suppress Suprathreshold Brightness and Colour Variations

“…However, we found here that texture contrast dominated colour contrast: 19% for Red vs. Texture, 14% for Violet vs. Texture. Further, Kingdom et al [29] found that red-cyan dominated violet-lime – though only by about 8% whereas here we find that texture contrast appears to be less dominant over violet contrast than it is over red contrast. It is clear that comparisons between any pair of cues cannot be inferred from their separate interactions with a third cue.…”

“…We now compare our results to those of Kingdom et al [28] , [29] . On average, in the Dark vs. Texture comparison our observers required 23% more luminance contrast in the Combined vs.…”

“…The above limitations notwithstanding, we suggest that the overall pattern of cue dominance (texture> colour> luminance) may reflect a hierarchy in which those cues that are more ambiguous with respect to material vs. illumination changes tend to be suppressed by those that are less ambiguous. Luminance is highly ambiguous in this regard, S-axis (violet-lime) colour variations less so, L-M (red-cyan) colour variations less still (hence Kingdom et al's, [29] result) and (we speculate somewhat) texture contrast the least ambiguous. In natural scenes luminance decrements and violet contrasts are more present in shadows than are red or texture contrasts [26] , and at least for patterned (vs. rough) textures, texture changes seem only to arise from material changes [11] .…”

“…Similarly we might expect some types of colour variation to dominate over others. For example, there is some evidence that red-cyan variations, which uniquely stimulate one of the two postreceptoral colour channels (see below) dominate over violet-lime variations, which stimulate the other colour channel [29] . It may not be merely coincidental that of the two types of colour variation, the red-cyan variations tend to be least contaminated by shadows and shading [26] , [30] , [31] .…”