TEXT AREA DETECTION ON COMPLICATED IMAGES

Andrianov A.I

230 KB

The methods of detection and locating the text areas on complicated images were researched. The requirements for the text recognition system for images contained complicated graphic scenes were determined. These images can have inhomogeneous background with no clear criteria’s to distinguish text and background. In addition these images can have various distortions. We have researched the applicability of MSER›s method for the solving the problem of selection the connected image areas. We propose to use AdaBoost’s and SVM’s classifiers for separation of the regions of the image as «textual» and «non-textual» segments. We use SWT algorithm as an additional attribute. It can be used for classifiers training, as well as for merging of probable-letters to words or to the lines of a text.

Библиографическая ссылка

Андрианов А.И ЛОКАЛИЗАЦИЯ ТЕКСТА НА ИЗОБРАЖЕНИЯХ СЛОЖНЫХ ГРАФИЧЕСКИХ СЦЕН // Научное обозрение. Физико-математические науки . 2020. № 1. С. 14-14;
URL: https://physics-mathematics.ru/en/article/view?id=3 (дата обращения: 24.06.2026).