DocumentCode :
3021664
Title :
Text recognition of low-resolution document images
Author :
Jacobs, Charles ; Simard, Patrice Y. ; Viola, Paul ; Rinker, James
fYear :
2005
fDate :
29 Aug.-1 Sept. 2005
Firstpage :
695
Abstract :
Cheap and versatile cameras make it possible to easily and quickly capture a wide variety of documents. However, low resolution cameras present a challenge to OCR because it is virtually impossible to do character segmentation independently from recognition. In this paper we solve these problems simultaneously by applying methods borrowed from cursive handwriting recognition. To achieve maximum robustness, we use a machine learning approach based on a convolutional neural network. When our system is combined with a language model using dynamic programming, the overall performance is in the vicinity of 80-95% word accuracy on pages captured with a 1024×768 webcam and 10-point text.
Keywords :
character recognition; document image processing; dynamic programming; image segmentation; learning (artificial intelligence); neural nets; character segmentation; convolutional neural network; cursive handwriting recognition; dynamic programming; language model; low-resolution document images; machine learning; text recognition; Cameras; Character recognition; Dynamic programming; Handwriting recognition; Image segmentation; Machine learning; Neural networks; Optical character recognition software; Robustness; Text recognition;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Document Analysis and Recognition, 2005. Proceedings. Eighth International Conference on
ISSN :
1520-5263
Print_ISBN :
0-7695-2420-6
Type :
conf
DOI :
10.1109/ICDAR.2005.233
Filename :
1575634
Link To Document :
بازگشت