A fuzzy based handwriting extraction technique for handwritten document preprocessing

In handwritten document image processing, a handwritten document is scanned by a flatbed scanner or camera arrangement. Ideally, we require to retain only the image of the handwriting without the background. Unfortunately, in many cases, background detail is also present in the scanned image. There are several factors contributing to this. One factor is the nature of the paper; it may contain color, pictures or patterns, it may be thin so handwriting on the reverse side shows through. Another factor is the handwriting implement used to produce the handwriting. Other possible factors can be defects of the scanner and improper setting of the scanning parameters. A fuzzy based image enhancement technique to extract just the gray scale handwriting from a handwritten document image is presented. The technique is based on fuzzy if-then rules applied on two-fuzzy sets whose parameters are estimated. The technique has been tried successfully on images of handwriting with bright-colored handwriting media, color paper, very thin paper, and paper with colorful pattern. The technique is particularly useful to the application of document processing that deal with gray level images.