Vision Sciences Society 2006
Volume 6, Number 6, Abstract 895, Page 895a doi:10.1167/6.6.895 ISSN 1534-7362

Spatial frequency integration for stereo processing in macaque visual area V4

Hironori Kumano Graduate School of Engineering Science, Osaka University, Japan
Seiji Tanabe Graduate School of Frontier Biosciences, Osaka University, Japan
Ichiro Fujita Graduate School of Frontier Biosciences, Osaka University, Japan


Early in the stereo processing system, neurons compute the disparity energy of stereo images filtered by their receptive fields. While filtering is important for the initial encoding of binocular disparity, it is not sufficient to account for many aspects of stereopsis. This discrepancy is reduced if disparity energy signals are integrated across filters of multiple spatial scales. Here, we studied the convergence of spatial frequency channels in cortical area V4 of two awake, fixating monkeys. We first measured the spatial frequency tuning of neuronal responses to sinusoidal gratings or two-dimensional filtered noise images. We then examined the disparity tuning with both correlated and anti-correlated dynamic random-dot stereograms (RDSs) for each neuron. Neurons with broader spatial frequency tuning had more attenuated disparity tuning for anti-correlated RDSs. In a subset of V4 neurons, we analyzed responses to various combinations of binocular disparity and spatial frequency by using two-dimensional filtered noise stereograms. The preferred disparity of most V4 neurons was consistent across a wide range of spatial frequencies. The independence of preferred disparity to the spatial frequency of the stimulus critically differs from the prediction by the standard disparity energy model that have a phase disparity between their receptive fields in the left and right eyes. We propose that V4 neurons pool disparity energy signals across spatial frequency channels, and this spatial frequency convergence contributes to an unambiguous representation of stereoscopic depth.
Supported by grants from MEXT (17022025) and Takeda Science Foundation.