A new simple and fast neural network-based algorithm for solving the correspondence problem is presented. The construction of the model was the result of i) redefining the correspondence problem, ii) redefining the functionality of simple cells, iii) an attempt to make a biologically plausible model, iiii) fusing a number of earlier models. An implementation of the model and some simulations are presented. The results are encouraging and shows that the model is fast, insensitive to noise and in severa aspects coherent with the behaviour of human stereopsis.
展开▼
