Assessment of thresholding algorithms for document processing

The thresholding technique used can critically affect the performance of subsequent operations such as page segmentation and character recognition. A taxonomy of thresholding methods has been proposed where the major categories are listed as entropy-based, histogram shape-based, object attribute-based, clustering based, adaptive, and object similarity-based. Furthermore a comprehensive performance evaluation of thresholding algorithms in the context of document analysis and character recognition systems has been performed. A large class of thresholding algorithms have been comparatively evaluated using shape distortion, false alarm and missprobability, and edge discrepancy measures. Both simulated documents and bitmaps of ground-truthed documents are used