My research interests contain image processing, information retrieval, and algorithm. Currently, I am mainly engaged in developping the computer systems for the document images that are hardly machine-readable, such as historical brush-written documents.
The product “document image retrieval system”, the word spotting engine applicable via the Internet, is published at http://records.c.fun.ac.jp/.
The further information is available at http://www.fun.ac.jp/~kterasaw/.
Attractive Factors of My Research
For the study of computer science, all I need is just a computer. I can try my ideas freely and can get a result immediately. Get an idea, try it, and be happy or unhappy depending on the result — isn’ t it a pleasure?
The online word spotting engine is published for three applications. First one is the website http://records.c.fun.ac.jp/. Second one is the commercial software released by a publisher. Last one is the web service released by another company. In addition, we are making collaborating with humanity researches and trying to apply information technology to the study of humanities.
The paper “Approximate Nearest Neighbor Search for a Dataset of Normalized Vectors” recieved the IEICE Best Paper Award 2009.
Major Books and Papers
- Kengo Terasawa and Yuzuru Tanaka
“Approximate Nearest Neighbor Search for a Dataset of Normalized Vectors”
IEICE Transactions on Information and Systems, vol.E92-D, No.9, pp.1609-1619, Sep. 2009.
* This paper won IEICE Best Paper Award on March 22, 2010.
- Kengo Terasawa and Yuzuru Tanaka,
“Locality Sensitive Pseudo-Code for Document Images,”
International Conference on Documents Analysis and Recognition, ICDAR2007, vol.1, pp.73-77, Curitiba, Brazil, Sept.23-26, 2007.
- Kengo Terasawa, Takeshi Nagasaki, and Toshio Kawashima, “Error Evaluation of Scale-Invariant Local Descriptor and Its Application to Image Indexing,” Electronics and Communications in Japan (Part III: Fundamental Electronic Science), vol. 90, no. 2, pp.31-39, Oct. 2006.）