A Comprehensive Study on Text Detection in Images and Videos

The appearance of a textual data in an images and videos provides the rich semantic information. Hence, the separation of the textual information from the image is an interesting and challenging task in computer vision and machine learning field. Text detection is a first step and foundation step in text extraction procedure, this step determines the existence of text pixels in an image. This paper provides the comprehensive study on various text detection methods. In this paper, the existing methods are categorized into caption text detection methods, scene text detection approaches and Hybrid text detection methods. Caption text detection methods works well for high contrasted images, because caption text gives clear edge information. Scene text detection methods works well for all challenges like complex back ground, low resolution, illumination and different orientated images. The scene text detection methods are developed based on pixel distribution concept and statistical approaches. Hybrid methods are the combination of both caption and scene text detection methods. Hybrid approaches are able to detect the text appears in several kinds challenges. Finally, tables demonstrate the observation of few important text detection methods with strength and weakness.