A Robust Text Localization Method for Natural Images

نویسندگان
Aminollah Mahabadi
Alireza Zarei
Department of Electrical Engineering, Shahed University, Tehran, Iran
چکیده
 Text localization from natural scene images is a challenging problem because of its complex backgrounds and large variations’ text patterns. In this paper, a multilingual text localization method based on connected components from natural digital images with a novel block density metric is proposed to detect and locate texts in natural Images.The method copes non-uniform background and different languages, fonts, colors, sizes, scales, distortion and nonuniform illumination within a tolerable elapsed time. The presented algorithm is experimented on varieties of two standard data sets and two additional data sets of camera and video images with Persian text. The experimental results show that the method proposed 90% recall, 85% precision, 87% f and MDR 13% of text-line level and 70% recall, 74% precision and 71% f of text-word level to achieve state-of-the-art performance. 

کلیدواژه‌ها

علوم رایانش و فناوری اطلاعات
دوره 15، شماره 1
بهار و تابستان
بهار 1396