Text Segmentation of Historical Degraded Handwritten Documents

Abstract:
The use of digital images of handwritten historical documents has increased
in recent years. These images provide important information and are used for genealogical and historical research. However,the poor quality of the original documents sometimes makes these images difficult to read and process.
We propose a novel technique that combines an iterative Otsu technique and bilateral filtering to allow automatically thresholding(segmenting) of handwritten text images. Preliminary results show that this technique helps segment the text in historical documents better than traditional thresholding techniques.