论文部分内容阅读
文献［１］的子波域滤波算法是通过将相邻尺度的相关量Ｃｏｒ２（ｍ，ｎ）作归一化处理后直接与Ｗ（ｍ，ｎ）相比较来决定变换值Ｗ（ｍ，ｎ）的取舍，并以噪声在各尺度的方差σｍ作为终止迭代的标准．但在实际计算中，简单地这样做会在小尺度上过多保留由噪声引起的变换点，导致重构的信号有较多的“毛刺”．本文提出了判别系数ＣＯＦ及阈值系数ＴＨ的概念，并对抽取信号边缘点的步骤做了改变以使ＣＯＦ易于选取，从而弥补原算法的不足．仿真表明，改进后的算法滤波效果较原算法有很大的提高．
The wavelet domain filtering algorithm in [1] is to determine the transform value W (m, n) by normalizing the cor- relation Cor 2 (m, n) ), And the variance σm of noise at each scale is taken as the standard of termination iteration. However, in practical calculations, simply doing so will reserve too many noise-induced transition points on a small scale, resulting in more “glitches” on the reconstructed signal. In this paper, the concept of discriminant coefficient COF and threshold coefficient TH is proposed, and the steps of extracting the signal edge points are changed to make the COF easy to select, so as to make up for the deficiencies of the original algorithm. The simulation shows that the improved filtering algorithm has a great improvement over the original one.