On the use of Moravec operator for text detection in document images and video frames

In this paper, we have proposed an accurate and an efficient text detection model for document images and video frames. The proposed model is based on the Moravec operator to localize text in a document image and horizontal and vertical mask convolution is performed to identify the actual text blocks. In order to exhibit the performance of the proposed model, we have conducted experimentation on the standard ICDAR-2003 dataset and also on the video frames collected from internet. The text blocks of video frames possess properties such as varying fonts, colour, size, languages etc. The superiority of the proposed text detection model is demonstrated through experimental comparison and objective comparison with the recently proposed text detection algorithms.