next up previous
Next: The Model Neuron Up: Appendix - Details of Previous: Disparity Estimators

The Coherence Network

The noisy disparity estimates from the first three network layers are combined into a stable percept by coherence networks for every view direction \( (i,j) \). The coherence network consists of two layers, a primary coherence layer \( \mathcal{C}(i,j) \) doing coherence detection and a final read-out layer \( \mathcal{R}(i,j) \). A single coherence layer \( \mathcal{C}(i,j) \) with corresponding read-out layer \( \mathcal{R}(i,j) \) realizes one of the disparity stacks \( \mathcal{D}(i,j) \) of Fig. 8. All disparity stacks \( \mathcal{D}(i,j) \) operate independently from each other.