Staff-line removal with selectional auto-encoders
作者：
Highlights：
• Staff-line removal is an important preprocessing stage in Optical Music Recognition.
• We consider a machine learning approach, in contrast to image processing techniques.
• Our proposal involves a new type of convolutional auto-encoder.
• The model is trained to remove the lines of the staves patch by patch.
• Our results significantly outperforms the other state-of-art strategies considered.
摘要
•Staff-line removal is an important preprocessing stage in Optical Music Recognition.•We consider a machine learning approach, in contrast to image processing techniques.•Our proposal involves a new type of convolutional auto-encoder.•The model is trained to remove the lines of the staves patch by patch.•Our results significantly outperforms the other state-of-art strategies considered.
论文关键词：Staff-line removal,Optical music recognition,Auto-encoders,Convolutional networks
论文评审过程：Received 22 December 2016, Revised 13 May 2017, Accepted 4 July 2017, Available online 27 July 2017, Version of Record 28 July 2017.
论文官网地址：https://doi.org/10.1016/j.eswa.2017.07.002
