The Study On Subpixel Document Segmentation

With the development of digital technology, document images can be easily obtained from books or magazines by using portable digital camera, scanner etc. Though the original documents are generally binary, document images are grayscale. Document images of grayscales can be displayed on screen, or used in printed form, but they are not good for recognize. At the same time, printing these documents directly seems not comfortable for reading and much consuming of printer ink.Segmentation is a basic procedure which converts document image into a binary image. Owing to this processing, segmented image is not only convenient for reading but also suitable for the following data treatments.Sub-pixel technology has been used mainly for edge detection. In this paper, in order to improve the segmentation quality, we formulate a sub-pixel algorithm to preprocess the document images. Experiment results show that it can improve segmentation quality significantly.Otherwise, this paper does some researches on image segmentation for uneven lighting document. For example, another information measure method is used for adaptive window based segmentation, and segmentation methods' advantage is mingled together for document segmentation.