Query Pixel Guided Stroke Extraction with Model-Based Matching for Offline Handwritten Chinese Characters

作者：

Highlights：

• A comprehensive scheme for solving the stroke extraction problem for handwritten Chinese characters is proposed.

• Stroke extraction is formulated as a problem of pairing and connecting skeletonwise stroke segments which are adjacent to the same cross region.

• Stroke connection ambiguity is solved by matching with template characters by tree search.

• High precisions of stroke extraction have been achieved on freely written characters.

摘要

•A comprehensive scheme for solving the stroke extraction problem for handwritten Chinese characters is proposed.•Stroke extraction is formulated as a problem of pairing and connecting skeletonwise stroke segments which are adjacent to the same cross region.•Stroke connection ambiguity is solved by matching with template characters by tree search.•High precisions of stroke extraction have been achieved on freely written characters.

论文关键词：Stroke extraction,Conditional fully convolutional network,PathNet,Stroke matching,Tree search

论文评审过程：Received 3 June 2020, Revised 12 August 2021, Accepted 1 November 2021, Available online 6 November 2021, Version of Record 13 November 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108416