複数の空間周波数情報を用いた脳模倣型両眼視差検出システム

抄録

We designed a binocular vision system that computes binocular disparity in real time based on the computational model of the visual cortex. The system consists of two image sensors, field-programmable gate arrays (FPGAs), and a personal computer (PC). To improve the accuracy, we employed the binocular energy model with multiple spatial band pass filtering and cooperative algorithm. FPGAs were employed to perform spatial filtering efficiently whose computational cost is high for conventional computers. The PC computes the disparity using the images pre-processed by the FPGAs. We examined the performance of our system by placing multiple objects at different depth levels in front of the system. The system provided a disparity map representing appropriate depth levels of the objects.