Text detection from natural scene images for Manipuri Meetei Mayek script

In this paper, we have developed a Text Detection System for Manipuri Meetei Mayek Optical Character Recognition (OCR). As an attempt to develop the complete OCR system for Meetei Mayek for natural scene and document images, this paper presents the prerequisite Manipuri Meetei Mayek script text detection part from natural scene images. A novel Text Detection algorithm is proposed using Edge Enhanced Maximally Stable Extremal Regions (MSER) and a new Stroke Width Transform (SWT) algorithm. The proposed algorithm has proved to be robust in detecting text and is the first Text Detection System reported for Meetei Mayek Script in literature. The segmented text from the natural scene images can be directly fed as input for feature extraction after skew correction if needed for developing the Meetei Mayek OCR.