Title :
Bayesian Best-First Search for Pattern Recognition - Application to Address Recognition
Author :
Hamamura, Tomoyuki ; Akagi, Takuma ; Irie, Bunpei
Author_Institution :
Toshiba Corp., Tokyo, Japan
Abstract :
In this paper, we propose a novel algorithm ldquoBayesian Best-First Search (BB Search)rdquo, for use in search problems in pattern recognition, such as address recognition.BB search uses ldquoa posteriorirdquo probability for the evaluation value in best-first search. BB search is more flexible to changing time limits compared to beam search used in conventional pattern recognition approach. It does not need designing a heuristic function for each problem like A* search.We demonstrated a 12.4% improvement over beam search on an address recognition experiment using real postal images.
Keywords :
Bayes methods; pattern recognition; probability; search problems; Bayesian best-first search; heuristic function; pattern recognition; search problem; Bayesian methods; Cities and towns; Databases; Detectors; Face detection; Face recognition; Pattern recognition; Search problems; Text analysis; Tree data structures; Bayes´ decision rule; a posteriori probability; address recognition; pattern recognition; search;
Conference_Titel :
Document Analysis and Recognition, 2009. ICDAR '09. 10th International Conference on
Conference_Location :
Barcelona
Print_ISBN :
978-1-4244-4500-4
Electronic_ISBN :
1520-5363
DOI :
10.1109/ICDAR.2009.231
