Extraction and Recognition of Open Captions Superimposed on TV News Articles
Yasuo Ariki · Asian Conference on Computer Vision · 2000
The purpose of this study is to automatically extract and recognize open caption characters superimposed on TV news. It starts from the extraction of stable frame sections including the open caption characters. Next process is the character region extraction using local line density to discriminate the characters from other elements such as lines and symbols. Final process is the character binarization by the floating adaptive three level thresholding (FATLT) which thresholds the image intensity into three levels and finally binarizes the character regions precisely even in low contrast, taking topological relation between characters and their background into consideration. We expanded our character extraction system in order to be able to recognize the characters and evaluated the character recognition accuracy.