Unsupervised Evaluation Methods Based on Local Gray-Intensity Variances for Binarization of Historical Documents

Abstract--We attempt to evaluate the efficacy of six unsupervised evaluation method to tune Sauvola's threshold in optical character recognition (OCR) applications. We propose local implementations of well-known measures based on grayintensity variances. Additionally, we derive four new measures from them using the unbiased variance estimator and grayintensity logarithms. In our experiment, we selected the well binarized images, according each measure, and computed the accuracy of the recognized text of each. The results show that the weighted and uniform variance (using logarithms) are suitable measures for OCR applications. 1

• Type: Conference
• Year: 2010
• Where: ICPR
• Authors: Marte Alejandro Ramírez-Ortegón, Raul Rojas
• Keywords: Computer Vision, ICPR 2010, Optical Character Recognition, Unbiased Variance Estimator, Unsupervised Evaluation Method