Gregor Žerovnik (2010) Image Database of Texts in Natural Scenes. EngD thesis.
There are many image databases, but the key part is having them well structured and organized. A well formed databse assembly methodology can take care of exactly that. In this diploma thesis I enlightned the databases from the computer vision area and their methodologies and created our own image database called CVL OCR DB. In the introductory part I briefly described the basics of databases and methodology. In the middle part I presented the most important image databases from the majority of computer vision subareas with the emphasis on the optical character recognition (OCR) databases. After their description I also compared them between each other. After that I focused on the methodology of the chosen databases, which were also described and compared between each other. Following that I presented our own database CVL OCR DB, which was compared to other OCR databases. The final part focuses mostly on the usage ilustration, analysis of results and on the future work.
Actions (login required)