Document image processing with stroke preservation and background suppression

Abstract

A method for improved readability of digitally captured document images that include printed or handwritten textual material overlaid on background graphics. Text in the document is generally composed at least in part of strokes characterized by pairs of edges in close proximity to one another. The digital image comprises an array of pixels representing at least a portion of the document. The edges in the pixel array are first identified and then grown by an effective amount so as to cause pairs of edges characterizing each stroke substantially to merge. The grown edges are then shrunk by an effective amount so as to cause non-stroke edges to be substantially eliminated without eliminating the merged stroke edges. A binarized representation of the merged stroke edges is then provided using an algorithm that is adapted to give a good representation of edge images, and a binarized representation of non-stroke regions of the pixel array is provided using a different algorithm especially adapted to represent the non-stroke regions. A document image with reconstructed textual matter may also be combined with a representation of the background graphics in which the overall contrast of the background graphics has been reduced. This may be achieved, for example, by scaling back the contrast of the captured digital image by a fixed percentage or by a variable amount prescribed by a formula, and then subjecting the scaled-back image data to a process such as a digital half-tone process providing a good binarized representation of the reduced-contrast image.