Handwritten Image Binarization Based on Background Estimation and Local Adaptive Integration

Abstract

In handwritten document images,there are often uneven lighting,ink stains,paper degradation,shadows and other complex conditions.In order to solve the problem that OCR effect of document image is not ideal after binarization in complex background,a binarization method of improved background estimation and local adaptive integration is proposed.In this method,the local adaptive binarization method is first used to obtain the binarization image with high recall rate,then the improved background estimation method is used to obtain the binarization image with high accuracy rate,and finally the two types of binarization images are integrated based on the connected domain method to obtain the final binarization image.Experimental results on DIBCO2013 and DIBCO2016 handwritten data sets show that the proposed method has better overall performance than Otsu,Wolf,Niblack,Sauvola,Singh and Howe.

Keywords

• image processing|handwritten document image|binarization|background estimate|binary integration