Neural network approach to the nonlinear shape restorations

Proposes a neural network approach to nonlinear shape restoration which is efficient regardless of the availability of the distortion models. Nonlinear mapping is extracted from the distorted image by using a reinforced learning SOFM(self-organizing feature map). For the exact extraction of the mappings between the distorted image and the original one, we define a disorder index in the SOFM, and we used this index to reinforce the training of the mappings selectively. Simulations are conducted on various kinds of distorted images with or without distortion models, and the results show that the proposed approach is very efficient and practical in nonlinear shape restorations.