Cooperation between Patterns Recognition and Contextual Analysis: Application to the Reading of Printed Documents
Journal of Machine Vision and Applications(1996)
摘要
This paper presents a cooperation between the recognition stage and the contextual analysis stage for printed documents recognition which makes contextual correction easier and the continuous training of the recognition stage possible. In the first part we recall the particular organization of our implementation which is based on a preclassrfication of identical characters patterns called prototypes. The second part describes the information provided by the recognition stage to the contextual stage in order to improve the syntactic correction. This information is represented by an ct initial recognition table w which gives all the hypotheses about the recognition of characters associated with a confidence rate computed by the recognition stage. We also demonstrate that the utilization of this table is more eficient for a syntactic stage using the Viterbi algorithm than the substitution table generally computed in advance on a large collection of characters. The third part describes a backloop from the contextual analysis stage towards the recognition stage. This will aNow the modrfication ofthe decision rules and the correction of the interpretation. It constitures the last link of the information chain that forms the cooperation between the context and the recognition.
更多查看译文
关键词
contextual analysis,pattern recognition,viterbi algorithm,decision rule
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络