Optical flow estimation by using the artificial neural network under multi-layers

A Hopfield model for computing optical flow is presented. A set of features describing the local intensity structure along the principal directions is used to measure the matching between the two local neighborhoods in the successive frames. The energy function can be derived based on the match measure and regularized by adding the Tikhonov stabilizer of the smoothness constraints. This energy function can be mapped onto an artificial neural network. The interconnection strengths between neurons and the bias inputs of the net can then be obtained. The synchronous scheme is used to determine the state change in the iteration procedure. This iterative procedure could be improved significantly by using the multilayer neural network. A smooth optical flow field with subpixel accuracy is obtained with only a few iterations at each layer and the final result is less sensitive to the noise distortion in the input image sequence than that by the conventional method