Regional Localization And Segmentation Of Connected Characters From Offline Handwritten Uyghur Images

With the development and popularization of the Internet, how to search information in the dynamic interest, word stream, shielding harmful information is the focus of academic and enterprises of great concern.Now there are two kinds of information on the network expression:(1)The machines can identify the ASCII text.(2)Image format.To extract the interested in words in the text, word is relatively easy.However, in the processing of handwritten text images is much more difficult。Because in dealing with a text image involves image processing, image matching,character segmentation, pattern recognition, and related technology.Especially dealing with handwriting text image.Due to the randomness and uniqueness of handwritten character segmentation, resulting in image matching difficulty.This paper will appear in the text in the images of Uyghur handwritten word extraction problem has taken the following solution.Handwriting handwritten Uyghur text in the image due to randomness, There will be a adhesion characters, so that word spotting from text images appear disturb. Hence the need for segmentation at great length the handwritten text in the image of character positioning and handwritten rules in writing under the condition of no damage.According to the gray smearing algorithm to find the positioning line handwritten Uyghur text in images,Then according to the touching characters positioning the concept of image connected domain,Finally through the search for the best segmentation point positioning line segmentation.To extract interested the words have segmentation,Due to the uniqueness of the handwritten character input from the computer in question is printed and handwritten character images is handwritten characters.So when the match due to the print characters compared with handwritten characters are different,the match will be difficult。So in this paper, the text first image is extracted and then all of the words segmentation,the BP model is set up for training,The input of printed characters as the input of the BP model for training, to print characters into handwritten characters,then calculate the matching image and matching image Harris corner points as feature points matching.Through this method,regional localization and segmentation of handwritten characters havegood results.