Fast Lexicon-Based Scene Text Recognition with Sparse Belief Propagation

Proceedings of the International Conference on Document Analysis and RecognitionPublished 1 September 2007

Jerod Weinman, Jerod Weinman, Erik Learned-Miller, Erik Learned-Miller, Andrew J. Hanson, Andrew J. Hanson

DOI10.1109/icdar.2007.4377061

Citations15

TL;DR

A flexible probabilistic model for character recognition that integrates local language properties, such as bigrams, with lexical decision, having open and closed vocabulary modes that operate simultaneously is proposed.

Abstract

Using a lexicon can often improve character recognition under challenging conditions, such as poor image quality or unusual fonts. We propose a flexible probabilistic model for character recognition that integrates local language properties, such as bigrams, with lexical decision, having open and closed vocabulary modes that operate simultaneously. Lexical processing is accelerated by performing inference with sparse belief propagation, a bottom-up method for hypothesis pruning. We give experimental results on recognizing text from images of signs in outdoor scenes. Incorporating the lexicon reduces word recognition error by 42% and sparse belief propagation reduces the number of lexicon words considered by 97%.

Keywords

Computer Science

ScholarlyCommons (University of Pennsylvania)Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data

12,978 Citations2001John Lafferty, Andrew McCallum +1 more

This work presents iterative parameter estimation algorithms for conditional random fields and compares the performance of the resulting models to HMMs and MEMMs on synthetic and natural-language data.

View PDF

IEEE Transactions on Information TheoryFactor graphs and the sum-product algorithm

6,481 Citations2001Frank R. Kschischang, Brendan J. Frey +1 more

A generic message-passing algorithm, the sum-product algorithm, that operates in a factor graph, that computes-either exactly or approximately-various marginal functions derived from the global function.

arXiv (Cornell University)Piecewise Training for Undirected Models

146 Citations2012Charles Sutton, Andrew McCallum

View PDF

Sparse Forward-Backward Using Minimum Divergence Beams for Fast Training Of Conditional Random Fields

55 Citations2006C. Pal, Charles Sutton +1 more

This work introduces sparse forward-backward, a variational perspective on beam methods that uses an approximating mixture of Kronecker delta functions, and motivates a novel minimum-divergence beam criterion based on minimizing KL divergence between the respective marginal distributions.

View PDF

Text recognition of low-resolution document images

44 Citations2005Chuck Jacobs, P. Simard +2 more

This paper uses a machine learning approach based on a convolutional neural network to achieve maximum robustness in OCR, and when 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/spl times/768 webcam and 10-point text.

EURASIP Journal on Advances in Signal ProcessingAn Embedded Application for Degraded Text Recognition

32 Citations2005Céline Thillou, Silvio Ferreira +1 more

View PDF

A Bayesian framework for fusing multiple word knowledge models in videotext recognition

20 Citations2003DongQuin Zhang, Shih‐Fu Chang

A formal Bayesian framework for videotext recognition by combining multiple knowledge using mixture models, and a learning approach based on Expectation-Maximization (EM) is presented in order to handle unseen words.

Improving Recognition of Novel Input with Similarity

17 Citations2006Jerod Weinman, Erik Learned-Miller

A probabilistic framework that unifies similarity with prior identity and contextual information is proposed that fusing information sources in a single model to eliminate unrecoverable errors that result from processing the information in separate stages and improve overall accuracy.

View PDF

Fast script word recognition with very large vocabulary

13 Citations2005Marc-Peter Schambach

For an HMM-based script word recognition system an algorithm for fast processing of large lexica is presented, where the computational effort for recognition itself can be reduced in the first step, while preserving recognition accuracy by the use of detailed information in the second step.

Fast lexicon-based word recognition in noisy index card images

13 Citations2004Simon M. Lucas, G. Patoulas +1 more

A complete system for reading type-written lexicon words in noisy images - in this case museum index cards - using a priority-queue based algorithm for the best matches with a set of words.