A wavelet neural network approach for shape from shading

A new approach for measuring the shape and surface of an object observed from a single view is proposed. The proposed approach is based on using a single hidden layer wavelet neural network (WNN) to approximate complex nonlinear reflectance map functions in which example data are not explicitly given. The shape from shading problem is formulated as the minimization of an intensity error function with respect to the network weights. Due to the properties of the WNN, the representation of network topology can be definitely developed, and the training problem can be transformed into a convex optimization process, the global minimum can be obtained and the learning speed also increases. Experimental results on synthetic images demonstrate the new algorithm has performed well and compared favorably to the traditional ones.