Texture Segmentation Performance as a Function of Target Size: Psychophysical Data and Simulation Results

Abstract

Texture segmentation performance is little affected by the size of the target patch if targets of 0.76 deg × 0.76 deg and 2.16 deg × 2.16 deg are used (Joffe and Scialfa, 1995 Spatial Vision 9 325 – 342). In two experiments we reinspected the results of Joffe and Scialfa, but now also included a rather small target patch (target sizes: 0.3 deg × 0.3 deg, 1.4 deg × 1.4 deg, and 2.3 deg × 2.3 deg). In agreement with Joffe and Scialfa, we found only a small difference in segmentation performance between the two large target sizes (1.4 deg × 1.4 deg and 2.3 deg × 2.3 deg) but a marked difference between 0.3 deg × 0.3 deg and 1.4 deg × 1.4 deg. In line with Joffe and Scialfa, we conclude that segmentation performance is affected by the size of the target in a nonlinear manner, leading to an asymptotic performance level at large target sizes.

In a simulation study we asked whether the observed experimental results could be explained within the framework of a standard two-layer ‘back-pocket’ spatial filter model. The results of the computations showed a good fit with the experimental data if spatial filters in the second layer of the model were set to a width of about 1 deg.