Staff lines removal based on naive bayesian model

In order to segment the complete musical notation, the usual step is to detect and delete the staff lines in the field of optical music recognition (OMR). Musical images are often affected by different strains, and it is difficult to delete the staff lines completely without affecting the integrity of the notes. A novel bayesian based method is proposed in this paper to solve the problem of staff lines deletion. We believe that staff lines have strong structural characteristics, and By learning its structure, staff lines can be deleted very well. Our algorithm first defines the appropriate size of the neighborhood window, and then the image operators constrained by this window is designed based on the bayesian model. The image operators can learn the structural features of staff lines, and it can be used as the convolution kernel to convolve with the image to obtain the image without staff lines. We have verified the feasibility of our algorithm on the data set 2013 ICDAR/GREC staff lines removal competition. Compared with other algorithms, the experiment shows that our algorithm can delete staff lines well.