Background Ruled-Lines Detection and Removal in Full-Colored Handwritten Image Documents

Abstract

Automation becomes the standard in nearly all aspects of life. Some of these aspects are text analysis, translating and retrieval. This requires machine typed format as a preprocessing step. Converting the handwritten text into machine printed counterpart requires Optical Character Recognition (OCR) system, which requires clean text as input. One of the problems facing the process of getting clean handwritten text is the ruled background lines which are intersecting and mixed with the text. In this work, we present fast algorithms for detection and removal of these ruled lines. The detection stage use only the centralized and squared part of the image document instead of wasting time if the whole image document is used, and use Hough transform for getting the ruled lines location and direction. The removal algorithm uses the color histogram segmentation for separating the text from the ruled lines. The Hue of the color is used to represent colors instead of using all color components. Then the segmented image document is morphologically enhanced and converted to binary image that is suitable for OCR. The results show the benefits of the proposed algorithms with F1-measures 91.43% and 88.52% for detection and removal respectively.