Detecting and recognizing LED dot matrix text in natural scene images

Abstract

This paper addresses a method for light-emitting diode (LED) dot matrix text detection and recognition in natural scene images. Unlike general text detection and recognition, the LED text detection is quite difficult to be done due to discontinuous character. In our proposed method, first, the Canny edge detector is applied to produce an edge image. From the edge image, the interesting points representing the center of a blob are extracted. These interesting points then are merged based on their properties to generate a character component. Through feature-based template matching, the filtering and recognizing process are performed simultaneously. Experimental results show that the proposed method is reliable, effective and fast to detect and recognize the LED text in natural scene images which general text method does not cover. © Springer-Verlag Berlin Heidelberg 2013.