Background Surface Estimation and Stroke Edge Detection by Using Document Image Binarization

Abstract

Badly degraded document images of text is very challenging task for segmentation .in this paper presents binarization technique ,to segment the text from badly degraded document images perfectly. this technique is based on text document usually have uniform color and texture. it has different intensity level. In the proposed technique, firstly background surface of document estimate through the polynomial smoothing procedure and compensated with different type of degraded document. then using compensated document detecting text stroke edge by using image gradient. finally segmenting text of degraded document by a local threshold, that estimated based on detected text stroke edge.