Text Region Detection Techniques for Natural Scene Images-A Review

Abstract

Abstract

The natural scene images tend to contain lot of information that can guide the persons navigating on the roads. Therefore understanding the text present in these images has become an area of active research. But the task of text region detection is a challenging task when text is to be detected from images that are captured from natural scenes. These images have varying font size, orientation, scripts, the text may be faded or blurred, the image may suffer from intensity in homogeneity and the text region may be occluded or may be missing. All these factors make the text region detection task difficult but interesting. Over the last two decades several methods for text region detection from natural scene images have been proposed, but till date no benchmark has been established that can accurately do this detection. The MSER based text region detection approach has been widely used, but it also fails if the text regions are blurred. In the presented work the various state of the art text region detection techniques proposed so far have been compared. The strengths and weaknesses of the techniques has been analyzed and based upon them the possible future directions has been proposed, so as to formulate a robust text region detection technique that overcomes the drawbacks of the existing techniques.

Keywords: Text Region Detection, Natural Scene Images, Survey