A scalable pattern spotting system for historical documents

作者：

Highlights：

• A new segmentation-free method for spotting patterns in document images is proposed.

• Our method is scalable thanks to the offline indexing and compact representation.

• We provide a unified benchmarking comparing our method with state of the art models.

• Experiments are provided on a dataset of medieval manuscript document images.

摘要

Highlights•A new segmentation-free method for spotting patterns in document images is proposed.•Our method is scalable thanks to the offline indexing and compact representation.•We provide a unified benchmarking comparing our method with state of the art models.•Experiments are provided on a dataset of medieval manuscript document images.

论文关键词：Pattern spotting,Word spotting,Image retrieval,Bag of visual words,VLAD,Fisher Vector,Product quantization,Historical document image analysis

论文评审过程：Received 15 May 2015, Revised 12 January 2016, Accepted 14 January 2016, Available online 2 February 2016, Version of Record 27 February 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.01.014