The performance evaluation of thresholding algorithms for optical character recognition

The paper presents performance evaluation of thresholding algorithms in the context of document analysis and character recognition systems. Several thresholding algorithms are comparatively evaluated on the basis of the original bitmaps of characters. Different distance measures such as Hausdorff, Jaccard, and Yule are used to measure the similarity between thresholded bitmaps and original bitmaps of characters.