Abstract
Both the local and global stages of motion processing are susceptible to noise in the stimulus. Given the hierarchical nature of motion processing, how do the effects of local and global noise interact with each other? If such interaction exists, does it differ across different types of motion patterns? We addressed these questions using a novel psychophysical technique, in which uncertainty at the global and local stages of motion processing was independently manipulated within the same motion stimulus. We used a multiple-aperture motion pattern, which consisted of an array of randomly-oriented, drifting-Gabor elements (Amano et al., 2009). Global noise was manipulated based on motion coherence: Global signal-to-noise ratio (global-SNR) was defined as the ratio between signal and noise element numbers. Signal elements were assigned velocities consistent with a specific global motion direction, while noise elements were assigned velocities based on random global motion directions. Local noise was introduced by superimposing dynamic-noise pixels on each drifting Gabor patch at every motion frame. Local-SNR was defined as the ratio between the contrasts of Gabor patches and noise pixels. Observers performed a 2-choice, global-direction-judgment task on three optic-flow patterns: translational (left vs right), circular (clockwise vs counterclockwise), and radial (inward vs outward). In each block of trials, we fixed local-SNR and measured the 75%-accuracy threshold in terms of global-SNR. For all three optic-flow patterns, we found a "tradeoff" between local and global noise: Global-SNR thresholds decreased log-linearly as local-SNR increased, suggesting an interaction between local and global noise in the motion system. Above a certain local-SNR level, global-SNR thresholds remained constant. This saturation point was lower for circular motion compared to radial and translational optic-flow patterns, suggesting that global integration mechanisms for circular motion are more tolerant to disturbances from local noise.
Meeting abstract presented at VSS 2016