TextProposals: A text-specific selective search algorithm for word spotting in the wild
作者：
Highlights：
• We present a text specific object proposals algorithm.
• Our algorithm is able to reach impressive recall rates with a few thousand proposals in different standard datasets.
• Our method generates word proposals without an explicit character segmentation.
• The combination of our object proposals with existing whole-word recognizers shows competitive performance in end-to-end word spotting, and, in some benchmarks, outperforms previously published results.
摘要
•We present a text specific object proposals algorithm.•Our algorithm is able to reach impressive recall rates with a few thousand proposals in different standard datasets.•Our method generates word proposals without an explicit character segmentation.•The combination of our object proposals with existing whole-word recognizers shows competitive performance in end-to-end word spotting, and, in some benchmarks, outperforms previously published results.
论文关键词：Object proposals,Scene text,Perceptual organization,Grouping
论文评审过程：Received 11 July 2016, Revised 1 February 2017, Accepted 28 April 2017, Available online 29 April 2017, Version of Record 11 May 2017.
论文官网地址：https://doi.org/10.1016/j.patcog.2017.04.027
