Scene text segmentation based on thresholding
Tutor / director / evaluatorŠarić, Matko
Document typeBachelor thesis
Rights accessOpen Access
This research deals with the problem of text segmentation in scene images. Introduction deals with the information contained in an image and the different properties that will be useful for image segmentation. After that, the process of extraction of textual information is explained step by step. Furthermore, the problem of scene text segmentation is described more precisely and an overview of more popular existing methods is given. Text segmentation method is created and implemented using C++ programming language with OpenCV library. Finally, algorithm is evaluated with images from ICDAR 2013 test dataset.