Neural Network for Invariant Image Classification
Prashanta Kumar Patra · IETE Journal of Research · 1996
In this paper, a Neural Network (NN) based approach for classification of images represented by translation, scale and rotation-invariant features is presented. The utilized network is a multi-layer perception (MLP) classifier with one hidden layer. The back-propagation learning is used for its training. The translation and scale invariant feautres are obtained by Fourier-Modified Direct Mellin Transform (F-MDMT). The rotation invariances are obtained by Zernike moments. Zernike moments are the mapping of the image onto a set of complex orthogonal polynomials. The performance of the MLP classifier is compared to three other traditional statistical classifiers, namely: Bayes, nearest-neighbour and minimum-mean-distance. Through extensive experiments with noiseless as well as noisy binary images of all Oriya characters (49 classes) and comparing these with usual geometrical invariants, it is reported that the neural network (MLP) has got a high degree of performance as a classifier.