Document Type : Research Paper
Abstract
Text detection in the videos is still considered an interesting and challenging problem for researchers in the field of computer vision and image processing. In this paper, we have been suggested a robust method for real-time text detection in the videos through determining the candidate text lines by using Laplacian edge detection and Morphological Dilation, then using an SVM classifier to eliminate the false candidate text lines. The experimental results for 6 different types of videos show that the proposed method able to detect text lines in real-time with a 98.2% recall rate and a 0.86% false alarm rate as well as it is able to detect different text line languages (English, Chinese, and Arabic ) with high accuracy of about 96.4%.