Design And Implementation Of Text Information Extraction On Smart Phone

Text is an important feature for many applications of computer vision. Text in camera based image carries useful information which is important for visual content understanding and retrieval. Text location is a main step of text extraction. The main purpose of text extraction is to transform text images into symbolic form, thus to modify, storage, retrieval, use and transmission. The key and difficult problem is how to locate text information fast and accurately. The complicated and changefully of nature scenes affects the accuracy of text location, and hardware resources for mainstream mobile devices also limit the transplant of some outstanding location algorithms(such as methods based on connected component, texture and edge, or their advantages' integration). With the development and popularity of multi-media technology and mobile devices, it has a great demand for the text analysis based on mobile devices. However, the image processing technology based on the smart mobile devices has no general frameworks. This work is an exploratory study of the text location and extraction on the Android mobile devices.This paper aims at the study of the natural scene analysis of the text based on smart ph ones on Android platform. In this paper, an image text location algorithm is implemented and improved and the text location in the image can be marked accurately and rapidly. We also propose an image binarization algorithm based on SWT( Stroke Width Transform). In our text location algorithm: firstly, we give a detailed implementation of Stroke Width Transform of input image and we improve the filtering process. The computation is linear in the number of edge pixels in the image and its calculation is simple. Secondly, Traditional grouping methods may fail to detected some Chinese words with complicated structures. We propose Chinese words grouping methods of structure analysis and improve the results. This paper also focus on the development of text location application based on the Android platform: firstly, take the hardware conditions of mobile devices into consideration, we utilize and modify some modules. With the OpenCV libs, we implement the core module of image text extraction system--text location module. Secondly, we utilize the MVC software design pattern and implement the text location terminal software based on the Android platform. Then the performance of the location algorithm is tested by ICDAR dataset and our dataset captured by the mobile phone. The design ideas and methods in this paper can give some useful suggestion to the field of text information analysis based on the mobile devices.