Statistical script independent word spotting in offline handwritten documents

作者：

Highlights：

• Line model allows one or more occurrences of keywords surrounded by filler models.

• We investigate five different filler models. CFMs provide the best results.

• The optimum number of character filler models varies for different scripts.

• Two background models are investigated to prune candidate keyword regions.

• Proposed framework is quicker and more accurate than current state of the art.

摘要

Highlights•Line model allows one or more occurrences of keywords surrounded by filler models.•We investigate five different filler models. CFMs provide the best results.•The optimum number of character filler models varies for different scripts.•Two background models are investigated to prune candidate keyword regions.•Proposed framework is quicker and more accurate than current state of the art.

论文关键词：Script independent,Keyword spotting,Hidden Markov models

论文评审过程：Available online 10 October 2013.

论文官网地址：https://doi.org/10.1016/j.patcog.2013.09.019