STEREO VISION USING A MICROCANONICAL MEAN-FIELD ANNEALING NEURAL-NETWORK

A simple stereo algorithm is presented here to obtain the 3D position of a scene's object points. The proposed algorithm can obtain an objec t point's 3D position by merging the microcanonical mean field anneali ng network (MCMFA) with the stereo vision system. Comparison with mean field annealing (MFA) or microcanonical simulated annealing (MCSA) re veals that current temperature controls the cooling speed, thereby red ucing the computation in MCMFA without degrading the performance. In a ddition, the initial temperature does not affect the quality of soluti on in MCMFA because of the new temperature cooling procedure. The corr espondence problem is the primary concern of stereo vision. A combinat orial optimization approach is used to resolve the correspondence prob lem for a set of features extracted from a stereo vision pair. An ener gy function is defined to represent the solution's constraints and the function is then mapped onto a 2D neural network. Each neuron in the network represents a possible correlation between a feature in the lef t image and one in the right image. The features are zero-crossing poi nts that are extracted using the LOG (Laplacian of the Gaussian) opera tor. Zero-crossing points are classified into 16 patterns according to their local connectivity. The difference of the sign value and direct ion value between a matched pair of zero-crossings can be used to set up the neural node's iteration rule. The network is assumed to be at i ts stable state when no change occurs in the neurons' state. Finally, the neighbour's disparity threshold (NDT) is used to enhance the preci sion of the corresponding situation. Once all the corresponding points are found, obtaining the 3D object position is a simple matter of tri angulation.