A neural network model in stereovision matching
Neural Networks
Intensity- and Gradient-Based Stereo Matching Using Hierarchical Gaussian Basis Functions
IEEE Transactions on Pattern Analysis and Machine Intelligence
Neural Networks: A Comprehensive Foundation
Neural Networks: A Comprehensive Foundation
The Geometry of Multiple Images: The Laws That Govern The Formation of Images of A Scene and Some of Their Applications
Non-parametric Local Transforms for Computing Visual Correspondence
ECCV '94 Proceedings of the Third European Conference-Volume II on Computer Vision - Volume II
| Hi-index | 0.00 |
A capability of depth perception in biological visual systems evolved throughout thousands of years to help animals and us, humans, to survive in a real life. This ability has helped us to navigate and avoid threatening obstacles. However, we still know very little about the biological processes that lead to such a perfection which is by far not achievable for artificial vision systems. Thus, proper models of these mechanisms would help in their better comprehension, as well as they could guide construction of better computer stereovision systems. In this paper we try to propose a new topology of an artificial neural network for the stereovision system. We try to construct a very simple model of a binocular system that is biologically inspired in a behavioral aspect and which, at the same time, is computationally efficient. It is a hybrid that consists of the convolutional, binocular receptive, and the Hamming neural networks. The input signal is non-parametrically transformed for better statistical preconditioning. The paper ends with the experimental results.