Bilkent University
Department of Computer Engineering


Textual Information Analysis on Multimedia Documents


Sercan Aksoy
MSc Student
Computer Engineering Department
Bilkent University

Textual information analysis on visual documents has become widely researched topic of data extraction, information retrieval and multimedia indexing, since dynamic text area detection is the fundamental step of text recognition systems. The existing state of art methodologies are able to provide convincing results on particular controlled datasets; however they might not be sufficient to be a generic solution that would work for both artificial (overlay, printed) and wild (text on the surfaces, handwritten text) scene texts. Our research approach is to obtain a generic algorithm for dynamic text area detection and recognition by investigating and improving state of art methodologies. As an initial consideration of our research, we studied on financial check images to practice OCR (Optical Character Recognition) algorithms for MICR (Magnetic Ink Character Recognition) purpose. In further steps, we have extended our perspective to dynamic text area detection and Word Spotting on Video Frames.


DATE: 23 March, 2015, Monday @ 15:40