Estimation ok a luminance nonlinearity in texture segregation

Purpose. Previously (Chubb el al.. .IOSA A II. 2350-2374. 1994), we have isolated a channel used for text me segregation and partial h characterized a luniinance nonhnearily in that channel (measuring Us 3rd- and higher-order polynomial components). We were unable to measure its 1st- and 2nd-order components (the lesponse to mean luminance and to contrast) due to intrusion by other channels with larger receptive lields. Here, we demonstrate a method for tull\ chaiacterizing this luniinance non linearity. Methods. We used the litnuion method of Chubb et al. Stimuli were alternating stripes of textures T ⁺ and T Fach texture was a sirid of uniform squares, with grey-levels v in T ⁺ and drawn from probability distributions (r) ±a(H9)A,ir ) + cos(6]A,(t ). there n is (he um form probability distribution. A, are /th-order orthogonal polynomials used as histogram unxlitiitfort. B is a direction in a space of miMincs of modulators J., and A and a is a hiiogmm cunirasl. For the pair i/,/1 = i 1.3) and a large set of directions 0, we determined the histogram contrast threshold a _θ foi detection ot the texture stripes (2AFC judgments of whether the stripes were oriented vertically or horiJ.onlally). Results. We report fits of a probability summation model to the data combining the previously isolated channel i respond nig to all polynomial orders) and an aggregate response ot any mhei channels (responding to only mean and contrast), allowing us lo estimate the lou-order components of the channel nonhnearil}, Conclusions. Histogram conliast analysis may be extended to fully characteri/e an early nonlincarity in the \iMial system.