A new image binarization technique by classifying document images

Abstract

The present work proposes a binarization algorithm based on classification of document images. The method first classifies the images into two categories namely, simple and complex images. The global threshold value is used for binarizing the simple document images whereas complex document images are binarized by applying local threshold values. A background checking method is introduced in this method to detect the blocks which can be marked as purely background blocks. Finally, a post-processing mechanism has been applied to improve the quality of the binarized image. © Springer-Verlag 2013.