Lobe Component Analysis Derives Disparity-Selective Filters

Neurons in the primary visual cortex are known to segregate in cluster with varying disparity preferences. They function as linear filters that help extracting depth information. This paper investigates the possibility to derive disparity-selective filters by lobe component analysis on stereo images. In experiments, we trained on eighty pairs of stereo images and tested on a sample object at various depths. Response curves were found to resemble characteristics expected for disparity neurons in the primary visual cortex