Android Based Application For Text Extraction And Recognition From Images

Our actual life scenes are full of various text images with high-level semantic information.How to extract text information of interest from these images has become the focus of researchers.Image text recognition technology includes text positioning,text segmentation,text recognition and other steps.But when using Android OCR software for text recognition,manual interception of text areas to achieve text area positioning is used,and now there are a lot of open source text recognition engine can be used in character recognition software based on Android.This enables the text recognition rate mainly depends on text image processing and character segmentation.To perform efficient image preprocessing and character segmentation of text images is very important while achieving the above steps.Text extraction is mainly divided into two parts:image preprocessing and text segmentation.This thesis is mainly focuses on obtaining text from the image after it has been located(images from books,indoor/outdoor prompts,restaurant menus,etc.).The text,complex background segmentation,character segmentation algorithm and its application in Android are studied in depth.The specific work accomplished as follows:(1)Image preprocessing:For the text image after positioning,the main purpose of image preprocessing is to realize the segmentation of the text and its complex background,which is an indispensable step in text extraction.The proposed method includes gray-scale,noise reduction,uneven illumination elimination,edge detection and a dual-threshold binary algorithm based on the Otsu method to achieve text segmentation.To a certain extent,the algorithm designed in this thesis can eliminate the interference of complex background and noise for text images with complex background and uneven illumination,so as to achieve the segmentation of target text and complex background;(2)Text segmentation:In order to improve the accuracy of text recognition,this thesis makes an in-depth study on the text segmentation algorithm aiming at the obtained high-quality text binarization image.In the specific character segmentation process,the horizontal projection method is firstly used for line segmentation,and then the vertical projection method is used to calculate the approximate width value of a single character in the text image(mainly targeting Chinese characters).At last,according to the approximate width of the characters,a variable template and a template change function are constructed to segment characters.To some extent,this algorithm overcomes the problems of wrong segmentation of left and right structural characters and conglutination characters,compared with the direct segmentation of characters by projection method.(3)Implementation of text recognition based on Android:In this thesis,the image character extraction algorithm is used as a technical means to design and implement Chinese character recognition software based on Android system.In addition to the text recognition function,the design of the software also adds real-time display functionality of Chinese character visualization image and voice playback function.