Nonlinear Preprocessing Based Document Image Binarization
Zhiyan Wang · Hainan Daxue xuebao. Ziran kexue ban · 2005
In many applications involving character recognition from digitized document images such as coupon and characters on chip, the input images are first binarized to form two-level images. However, direct binarization of such images often results in unsatisfactory performance because of the poor quality of the input images. Consequently, it is usual to preprocess such images in order to enhance the edges in the presence of noise and other types of distortions that occur during the scanning process. Edge enhancement and noise reduction are conflicting requirements for linear, space-invariant filters. This paper introduces a nonlinear method for preprocessing document images that can denoise and enhance the edges at the same time.