Text Image Orientation Research And Application Based On Text Structural Features

With the appearance of the OCR (Optical Character Recognition) technology, it is possible to transfer the content of paper documents into editable text. OCR technology has been relatively mature till now, the correct placement of text, the image captured by scanning or camera has a high recognition rate. Character recognition technology combined with the ordinary camera, to be used for voice broadcast on portable intelligent reader, to facilitate the visually impaired people, in order to obtain a broader knowledge and information for them. However, the direction of the image (ie, the direction of the text) plays a decisive role in the recognition results. When using the reader, the reader slot could avoid inclined paper, but the blind or visually impaired people could not correctly identify the direction, thus the reader practicality is greatly reduced. Therefore, to achieve the certain text direction (0°,90°,180°,270°) judgment and correction, is the main content of this paper. The effect of text determination and correct image orientation, have multiple aspects as influencing factors, such as typesetting of the text, text font, and whether it contains pictures, etc, which will affect the effect of the algorithm in varying degrees, and add the correction to the text direction with a certain complexity.The main contents of this paper are:image preprocessing, typesetting text judgment, corrections and application of text direction. Among them, the image preprocessing includes image gray processing, image binarization, noise removal and character segmentation; Text formatting judgments include projection theory, determination of the textline, horizontal and vertical text detection;Text Orientation Correction and application section includes feature extraction of strokes of Chinese characters, determination of the direction of the text correction of the text image orientation, and the experimental effects of this application. During the study of this subject, firstly, some image preprocessing algorithms are outlined, because pretreatment of the algorithm is directly related to the quality of the effect; then based on the results of pre-treatment, use the algorithm to achieve typesetting text detection. Finally, according to the Chinese character stroke feature extraction, determine the direction of the text, and ultimately achieve the purpose of the correction of the text direction.By comparing the experimental results and analysis, the data of accuracy and timeliness on aspects of the algorithm is obtained.In the experiment, the proposed algorithm has been used to analyze and correct the text image. Experimental results show that the proposed algorithm achieved text layout distinction and the Correction of text direction, which has a high accuracy, thus make it possible for blind reader to be used intellectively.