一种快速稳健的乐谱图像谱线检测与删除方法

提出一种快速稳健的乐谱图像谱线检测与删除方法。在谱线检测阶段,首先利用计算若干条形区域水平投影队列间交叉相关性的方法,自动矫正图像的倾斜和变形缺陷,然后根据水平投影峰值检测谱线。其优点是算法执行效率高,对有倾角和变形的图像具有良好的鲁棒性,图像经过矫正,谱线定位的精度得到显著提高。在谱线删除阶段,提出基于游程邻接图段分析的删除方法,避免了传统方法的“误删”现象发生,较好地保持了音符的完整性。实验结果表明了上述方法的正确性和有效性。

A Fast and Robust Approach to Detect and Remove Staff Lines from Music Score Image

Staff lines play an important role in Optical Music Recognition.Detection and removal of staff lines is a very key step before next processes of note symbols recognition.In the paper,A Fast and Robust Approach is introduced to detect and remove staff line from score image.At the stage of staff lines detection,horizontal projection of adjacent vertical image strips and cross-correlation algorithm is employed to correct the skewed and deformed image,and then the position of staff lines is located through finding peaks of the projection.At the stage of staff lines removal,a run-length adjacency graph of image is created first,and then the graph is segmented into some primitive sections with the same width and topology.Based on analyzing features of those primitive sections,staff lines can be removed without damaging symbols around staff lines.Experimental results show the presented approach can detect and remove staff lines effectively and robustly.