The present work addresses the field of automated detection of text elements in still and moving images. Text detecting algorithms act as the basis for various systems which include applications for automated video indexing, automated recognition and translation of information boards, text-to-speech software for the disabled, and others. A comprehensive overview of existing text information extraction (TIE) systems is presented including their similarities and differences. Furthermore, this thesis introduces the Text-Spotting Agent, an algorithm that can be integrated into any TIE-system to localize scene text in still images. The algorithm recognizes text by detecting typical edge-direction combinations appearing in characters. In addition, a prototypical Java implementation of the Text-Spotting Agent is described. Results from using the newly presented algorithm are also discussed.