Using character recognition and segmentation to tell computer from humans
Authors:
- P.Y. SimardPSP.Y. Simard
- R. SzeliskiRSR. Szeliski
- J. BenalohJBJ. Benaloh
- J. CouvreurJCJ. Couvreur
- I. CalinovICI. Calinov
- 2 March 2004
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 1, 418-423
Abstract
How do you tell a computer from a human? Thesituation arises often on the Internet, when online pollsare conducted, accounts are requested, undesired email isreceived, and chat-rooms are spammed. The approachwe use is to create a visual challenge that is easy forhumans but difficult for a computer. More specifically,our challenge is to recognize a string of randomdistorted characters. To pass the challenge, the subjectmust type in the correct corresponding ASCII string.From an OCR point of view, this problem is interestingbecause our goal is to use the vast amount ofaccumulated knowledge to defeat the state of the art OCRalgorithms. This is a role reversal from traditional OCRresearch.Unlike many other systems, our algorithm is based onthe assumption that segmentation is much more difficultthan recognition. Our image challenges present hardsegmentation problems that humans are particularly aptat solving. The technology is currently being used inMSN's Hotmail registration system, where it hassignificantly reduced daily registration rate with minimalConsumer Support impact.Keywords
This publication has 7 references indexed in Scilit:
- Rapid object detection using a boosted cascade of simple features2005
- PessimalPrint: a reverse Turing testInternational Journal on Document Analysis and Recognition (IJDAR), 2003
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998
- LeRec: A NN/HMM Hybrid for On-Line Handwriting RecognitionNeural Computation, 1995
- Dynamic planar warping for optical character recognition1992
- Fast algorithms for low-level visionIEEE Transactions on Pattern Analysis and Machine Intelligence, 1990
- Euclidean distance mappingComputer Graphics and Image Processing, 1980