基于小波变换尺度相关性的轮廓数据去噪

基金项目:

四川省科技攻关资助项目(04GG010-004-1)

详细信息

Contour Data Denoise Based on Inter-Scale Correlations of Contourlet Transform

1. School of Mechanical Engineering, Southwest Jiaotong University, Chengdu, 610031, China;
2. Institute of Applied Electrinics AEP, Mianyang, 621900, China

摘要: 针对机械零件CT切片轮廓数据存在大量噪声的特点,在总结小波变换方法的基础上,提出了基于曲率小波变换的多尺度分析方法,利用尺度间相关性达到去噪和特征点检测目的,不仅解决了单一尺度无法有效去除切片轮廓存在的冗余数据和噪声,而且由于线特征剧烈波动而导致轮廓特征点检测困难,拟合精度差等的问题也得到有效地解决。
- 曲率 /
- 小波变换 /
- 轮廓 /
- 相关性
Abstract: Because CT slice outline data has a lot of noise,, proposed the multi-scale analysis method based on the curvature wavelet transform. Using the inter-scale correlation to achieve denoise and features detection. Not only solved the single-scale can not effectively remove contour noise, but also solved the contour detection difficult and the poor fitting accuracy problems which were caused by sharp fluctuations of line features.
- curvature /
- wavelet transform /
- profile /
- correlation