Alpha-Shape Based Classification with Applications to Optical Character Recognition

We present a new classification engine based on the concept of alpha-shapes. Our technique is easy to implement and use, time-effective and generates good recognition results. We show how to efficiently use the concept of alpha-shapes of low dimension to support data in arbitrary dimension, thus overcoming the lack of shape algorithms in high dimensions. We further show how to elegantly choose suitable primitives to capture desirable shapes that tightly bound the data. We present experiments showing that our technique generates good results with Optical Character Recognition (OCR) tasks. Based also on strong theoretic properties, we believe that our technique can serve as a desirable classification engine for various domains in addition to OCR.