Abstract

Given that OCR accuracy is not yet adequate for handwritten scripts, specially with large lexicons, instead of indexing the OCR'ed text of handwriting for retrieval, various approaches based on other information from the document images have been proposed. In particular, information retrieval of handwritten documents has been dealt with in two ways: keyword spotting and document retrieval.
Keyword spotting methods can be performed by template-based or template-free approaches. The template-based approach is based on image matching and requires an image of the query word as a template. Therefore it is only applicable to constrained handwriting such as the handwriting of the same person or in case of small lexicons. In the template-free approach, the similarity between a word image and a query is measured by the recognition score given by a word recognizer. It is more robust for applications of …

Date

2014

Authors

Huaigu Cao, Venu Govindaraju

Book

Advances in Digital Document Processing and Retrieval

Pages

105-122