DETECTION AND REMOVAL OF BASE LINES FROM DOCUMENT IMAGES

Fast Hough transform is utilized to detect base lines from document images. Feature points taken from the image are divided into two subsets, from which two feature points are selected respectively to compute parameters in Hough space. A threshold is set such that any bin exceeding the threshold means the evidence of the reference line and the detection will stop. The proposed Hough transform can be implemented at fast speed with practical value. The base line is removed by using mathematical morphology operation, while the structural element is decided according to the width of the base line and the intersectional stroke. Character strokes are protected while the base line is removed. Experimental results with envelope images show that the presented method can detect and remove base lines effectively.