A Shape Based Text Segmentation Using Eccentricity Transform
Document images are stored, manipulated and accessed easily on an electronic media. There is an increasing
demand to design automated systems for the purpose of extracting useful information from document images. Most
documents, suffer a lot of challenges including occlusion to detect text. Also, techniques required to solve such challenges
demand more computational processing and falls short in terms of real-time application. In our project, we have made use of
eccentricity transform to segment the text into line(s), word(s) and character(s) making it efficient for video processing in
real-time application. We have further extended our work to process the algorithm on Raspberry Pi 2 board for video
processing. The algorithm has been verified and tested on 30 printed English documents.
Keywords— Document Images, Segmentation, Eccentricity Transform, Raspberry Pi 2.