Phase unwrapping in 3-D shape measurement using artificial neural networks

Two observations are worthy of note. First, the experienced optical engineer can usually determine, sometimes partly subjectively, the positions of phase wraps in the image. This suggests that the information necessary to identify phase wraps does exist. Second, to date, no universally applicable technique for phase wrap detection is available. Indeed, it may be that there is no straight forward analytical method that can be used. The concept of a neuron was first postulated by McCulloch and Pitts (1943) and a neural network provides a mechanism by which a machine can learn from experience. The foregoing discussion suggests that neural network technology may be suitable for addressing the phase unwrapping problem. This paper describes preliminary work on the use of neural networks to identify phase wraps earlier in the phase measuring process, prior to the calculation of wrapped phase