Staff removal using image operator learning

作者：

Highlights：

• We propose a learning based method for staff removal that does not impose any restriction on the images.

• We provide a reference parameter set for our method and use feature selection to determine parameters from training data.

• The method is shown to be robust both to parameter changes and to many types of deformations in the music scores.

• Learned operators' results are par with state-of-the-art heuristic methods and superior to many methods for staff removal.

摘要

Highlights•We propose a learning based method for staff removal that does not impose any restriction on the images.•We provide a reference parameter set for our method and use feature selection to determine parameters from training data.•The method is shown to be robust both to parameter changes and to many types of deformations in the music scores.•Learned operators' results are par with state-of-the-art heuristic methods and superior to many methods for staff removal.

论文关键词：Staff removal,Optical music recognition,Document image analysis,Image operator,Machine learning

论文评审过程：Received 26 February 2016, Revised 30 September 2016, Accepted 1 October 2016, Available online 4 October 2016, Version of Record 22 October 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.10.002