Abstract: Extraction of text from the images is one of the important and challenging tasks. These images can be captured from digital capturing devices, since most of the people use digital cameras and other devices wherein text information is important among all the contents from the images. This has many applications such as reading license plate, helping blind persons, tourists etc. In this paper, an approach to detect and locate the texts in images is presented. Text detection and extraction is hard due to the intricate background, different alignment of text size, orientation in images. Texture-based method and the features of wavelet are used to design the proposed work. In the pre-processing step, the image is converted and divided into 50×50 blocks and then extract texture features. Next the classifier classifies the blocks into text and non-text and finally, merging to have text regions. The proposed method accomplishes a detection rate of 99% on a variety of images.
Keywords: Text Detection, Text Extraction, Texture-based, Wavelet Features