Abstract
This paper deals with discriminating textural patterns of geometrically dot-arrayed figures by means of optical Fourier's transform.
A set of original patterns, which have idealized regularity and homogeneity as texture, is defined as that of categories to be recognized. In order to extract the parameter of feature, both the elementary shape of the pattern and the array of the element are introduced. A fundamental theory to recognize the input textural pattern as one of the category patterns is derived premising optical Fourier transform. Some experiments are carried out using an optical set-up by which the Fraunhofer diffraction of the input pattern is obtained. The input patterns used in the experiment are such that they are fluctuating as a whole pattern around a regular array fixed in the space. The experiments show that the method proposed in this paper is so effective that so much fluctuated textural patterns as being visually difficult to recognize are clearly recognizable.
