Automated Text Detection and Text-Line Construction in Natural Images
This work develops an automated system to detect texts in natural images captured by the cameras embedded on mobile devices. Unlike former researches which focus on detecting with straight texts, this work proposes a text-line construction algorithm which
- PDF / 363,382 Bytes
- 7 Pages / 439.37 x 666.142 pts Page_size
- 26 Downloads / 216 Views
Abstract This work develops an automated system to detect texts in natural images captured by the cameras embedded on mobile devices. Unlike former researches which focus on detecting with straight texts, this work proposes a textline construction algorithm which is able to extract curved text-lines in any orientations. An image operator called the Stroke Width Transform is adopted to exploit connected components which have stroke-like properties. Text components are classified into two types: active and passive. The links of active components are considered the initial orientation of text-lines. Complete text-lines are constructed by linking active and passive components. The system is implemented on the Android platform and the experimental results demonstrate the feasibility and validity of the proposed system. Keywords Text detection
Stroke width transform Mobile application
1 Introduction Employing text detection algorithms on mobile devices assists users in understanding or gathering useful information. Today, many advertisements embed specific QR-codes, allowing users to capture the code using the camera on C.-C. Yu (&) W.-H. Hsu T. C. Chuang Department of Computer Science Information Engineering, Vanung University, 32061 Zhongli, Taiwan, Republic of China) e-mail: [email protected] Y.-N. Chen Department of Computer Science Information Engineering, National Central University, 32001 Zhongli, Taiwan, Republic of China.)
S.-S. Yeo et al. (eds.), Computer Science and its Applications, Lecture Notes in Electrical Engineering 203, DOI: 10.1007/978-94-007-5699-1_68, Ó Springer Science+Business Media Dordrecht 2012
677
678
C.-C. Yu et al.
smartphones and then direct them to a specific webpage. Others may ask users to type in keywords on browsers. A system incorporating with text detection and Optical Character Recognition (OCR) techniques provides a more convenient, flexible, and intuitive approach for users. Mobile applications with text detection techniques can greatly assist people. For example, Ezaki et al. [1] proposed a textto-speech system to assist visually impaired people to understand the content of an image. Algorithms used for detecting texts are categorized into region-based method and connected component-based method. Region-based methods adopt features such as edges [2], DCT or wavelet coefficients [3], and histograms of oriented gradients [4] are used for analyzing whether the features in a sliding window reveal text-like properties. However, this kind of algorithm is not suitable for applications in mobile devices due to the high computational complexity. Another approach is the connected component-based method. Text candidates are extracted by thresholding firstly. Then, non-text components are filtered out based on certain geometric rules. The connected component-based method is usually computationally inexpensive than the region-based method. Therefore, applying a connected component-based method to mobile devices is a reasonable option. In terms of connected component-based
Data Loading...
