基于局部二进制模式的乐谱谱线检测与删除

发布时间：2018-04-23 23:10

  本文选题：谱线检测与删除 + 光学乐谱识别　； 参考：《计算机科学与探索》2017年12期

【摘要】：谱线检测与删除是光学乐谱识别中重要和关键的环节之一。在乐谱中,谱线往往与大多数符号交叉或重叠,即存在像素属于谱线像素同时也属于符号像素的情况,因此删除谱线并且不破坏音乐符号并非易事。研究目标是需要删除仅仅属于谱线的像素,观察乐谱图像可以发现谱线像素与非谱线像素局部纹理存在差异,主要表现为谱线像素的局部纹理与谱线宽度相关,简洁明了,而非谱线像素的局部纹理除了存在仅与自己本身相关的情况,还存在与交叉点相关的情况。因此,采用局部二进制模式通过提取局部纹理特征,获得谱线像素与非谱线像素局部纹理的差异,对谱线与非谱线像素进行检测分类,进而将谱线像素删除。该方法不仅可以删除理想状态下乐谱谱线,对弯曲状态下乐谱谱线同样适用。实验结果证明了该方法在像素误差、片段误差等性能指标上优于现有常用方法。
[Abstract]:Spectral line detection and deletion is one of the important and key links in optical music spectrum recognition. In music, spectral lines are often crossed or overlapped with most symbols, that is, pixels belong to spectral pixels as well as symbol pixels, so it is not easy to delete spectral lines without destroying music symbols. The goal of the study is to delete only the pixels belonging to the spectral lines. Observing the music spectrum images, we can find that there are differences in the local textures between the line pixels and the non-line pixels, which mainly show that the local textures of the line pixels are related to the spectral line width and are concise and clear. The local textures of non-spectral pixels are not only related to themselves, but also related to intersection points. Therefore, using local binary mode to extract local texture features, the difference between spectral line pixel and non-spectral line pixel is obtained, and the spectral line pixel is detected and classified, and then the spectral line pixel is deleted. This method can not only delete the music spectrum line in ideal state, but also apply to the music spectrum line in bending state. The experimental results show that the proposed method is superior to the existing methods in pixel error, segment error and other performance indexes.
【作者单位】： 天津大学电子信息工程学院;
【基金】：国家自然科学基金Nos.61101225,60802049,61471263 天津市自然科学基金No.16JCZDJC31100~~
【分类号】：J613.4;TP391.41
，

本文编号：1794054