PERFORMANCE ANALYSIS OF AN OCR SYSTEM VIA AN ARTIFICIAL HANDWRITTEN CHINESE CHARACTER GENERATOR

Abstract

A handwritten Chinese character generator that generates handwritten Chinese characters with different variations is proposed and used to evaluate the performance of an OCR system. The character generator first generates radicals of a character stroke by stroke and then combines the generated radicals to form a line-vector character. Next, the line-vector character is thickened to obtain a character image. Characters generated by this method will have great variance in shape but will still satisfy the structural constraints. The generated characters are then used to perform two types of evaluations. First, the stability of stroke extractors in terms of stroke number is evaluated. Two stroke extractors, one based on thinning and the other on vectorization, are analyzed. The vectorization-based stroke extractor we propose operates directly on the run length codes of line segments. From experimental results, we find that the thinning-based method is more time-consuming but more stable than the vectorization-based method. Second, the peak performance of a matching module is evaluated and the recognition error caused by stroke extractors is identified.