Optimum Circular Symmetrical Filters and Their Uses in Optical Pattern Recognition

Publisher Website
Google Scholar
Abstract
A new type of filter called the optimum circular symmetrical filter is presented and its uses in optical pattern recognition are described. The filters can be realized by optical or computer-generated holography. They are space-invariant and rotation-invariant, and are optimized in the mean-square sense. This makes the recognition process fully parallel, without input scanning or coordinate transformation, no matter where and in what orientation the patterns to be recognized are located in the input plane. Multiple use of those filters can further improve system recognition ability. By using them jointly with other methods of linear combinations of matched filters, more advanced systems can be obtained.