MICRO-OPTICAL MULTIWAVELET ELEMENT FOR HYBRID TEXTURE SEGMENTATION PROCESSOR

The concept of multiresolution analysis on the basis of multichannel f iltering in the theory of the early stages of human vision offers suit able methods for pattern recognition. A novel micro-optical multiwavel et element, which has the functions of wavelet filtering, beamsplittin g, and self-imaging, has been designed and fabricated for a hybrid tex ture segmentation processor. All sixteen channel filters process in pa rallel, using a conventional 2-D correlator. Accordingly, sixteen filt ered images sensitive to different scares and orientations are obtaine d simultaneously as features for texture segmentation. Experimental re sults and primary applications are presented. (C) 1998 Society of Phot o-Optical Instrumentation Engineers. [S0091-3286(98)02201-6].