2nd-order contrast discrimination

Purpose. Luminance contrast discrimination functions (increment threshold: ΔC as a function of C) are often dipper-shaped, with an initial improvement in threshold contrast increment followed by a region of near unity slope. The first region is attributed to subthreshold summation, an accelerating nonlinearity or uncertainty. I measured this function for 2nd-order, texture modulation contrast. Methods. Stimuli were texture patterns of the form T(x,y) = √1 + C m (x) V(x,y) + √1 - C m(x) H(x,y), where C is the 2nd-order contrast and H and V are independent horizontally- and vertically-oriented filtered noise textures (8 cpd peak frequency, 30° orientation bandwidth). The square root ensures that expected 1st-order contrast power is constant across the stimulus. The spatial sinusoidal modulator is m(x) = sin(2πfx + φ), where phase φ was randomized across stimuli. Subjects indicated which of two stimuli (of modulation contrasts C and C + ΔC) had higher modulation contrast (2IFC), Results. In previous work (Landy & Ternes, OSA '95), we found that threshold modulation contrast (threshold ΔC for C = 0) as a function of modulation frequency f was very broadband and scale-invariant. We now find that contrast discrimination has a dipper shape, but threshold barely emerges from the bottom of the dipper and never reaches the near unity slope region before the maximum physically achievable modulation contrast is reached. Conclusions. The same explanations of the dipper shape may apply to 2nd-order as well as 1st-order contrast discrimination.