Human efficiency for discrimination of 3-D symmetry

Abstract

We conducted an experiment to evaluate the statistical efficiency of human for discrimination between 3-D symmetry, i.e., the 3-D pattern symmetrical about a sagittal plane, and 3-D random which was non-symmetrical about a sagittal plane but symmetrical about a vertical axis in the fronto-parallel projection plane. We added Gaussian depth noise to the 3-D symmetry pattern and the 3-D random pattern. The discriminability (d′) of the observers was obtained, and was compared with that of the ideal cross correlator. In general, the efficiency was lower than that for 2-D symmetry discrimination, but similar to that for other 3-D tasks using dot stereograms. Moreover, the efficiency for 3-D symmetry discrimination was lower when the pattern was composed of more elements, and it decreased along with increment of the noise variance in longer duration condition (1500ms). From this result, we estimated the number of elements which observer sampled in performing the task.