A new approach to object classification in binary images

In this paper, we address the problem of classifying binary objects using a cascade of a binary directional filter bank (DFB) and a higher-order neural network (HONN). The binary DFB receives as input a binary image and returns as output a binary subband representation. Because processing is performed on a finite field, the DFB is able to operate efficiently. Furthermore, the DFB provides a representation that delineates the directional components in the image, which enables the HONN to exploit the underlying shape of the object effectively. The paper provides a description of the new binary DFB and its use with the HONN, all in the context of object classification