Topological invariants and their usage

Author: Giorgi Bakradze
Co-authors: Koba Gelashvili
Keywords: topological invariants, image processing, OCR
Annotation:

Seminar presents the results of research conducted in 2016 about alternative ways of optical character recognition. The core idea of our research was topological invariants. For each symbol we define unique topological characteristics to classify groups of symbols. From practical standpoint extracting invariants requires heavy preprocessing of source image, which is multi step process. One of the key steps during preprocessing is graph based thinning. Results obtained from this step is then used for extracting features mentioned above. Also for testing and experimentation we needed flexible framework for changing and adapting, so different design patterns were used and modified for our needs.

Lecture files:

Topological invariants and their usage [en]
სიმბოლოების ტოპოლოგიური ინვარიანტების განსაზღვრა და გამოყენება [ka]