-
摘要: 提出一种鲁棒性的谷脊线提取与增强算法. 算法采用多步逼近的策略: 首先根据每个点的局部最小二乘拟合曲面多项式计算每个点的主曲率, 并用绝对值较大的主曲率标识出谷脊潜在特征点; 然后通过将特征点投影到离其最近的潜在特征线上得到增强的特征点; 再对增强后的特征点进行平滑, 选择合适的平滑点生成特征折线; 最后再对特征线进行进一步的扰动滤除等操作得到光滑的谷脊线. 实验结果表明, 本文算法稳定、抗噪性强、能满足多分辨率的特征提取要求.Abstract: We present a robust algorithm for extracting valley-ridge lines from point set model. Our algorithm is based on multi-step refinement operations: Using moving least square method, we fit a smooth patch for neighborhood of each point, then calculate the principal curvature for each point. Potential valley-ridge points are identified according to the biggish principal curvature. After enhancing valley-ridge points by projecting onto their closest potential feature lines, we smooth the projected points and grow polylines through the smoothed points. Finally, smooth valleys and ridges are achieved by resolving the results. Experiments indicate that our algorithm has good stability and strong anti-noise performance, and it also leads to good results of multi-resolution extractions of features.
-
Key words:
- Point-set /
- valley-ridge extraction /
- feature enhancement /
- valley /
- ridge /
- moving least squares (MLS)
-
计量
- 文章访问数: 3050
- HTML全文浏览量: 107
- PDF下载量: 1690
- 被引次数: 0