一种子波域滤波算法的改进

Improvements on an Adaptive Filtering Algorithm in Wavelet Transform Domain

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．

In the filtering algorithm of Ref.,it is by directly comparing the absolute value of Corr 2( m,n ) after being rescaled and W(m,n) to determine whether the data point W(m,n) should be kept.And the standard deviation σ m of noise at each scale m is employed to stop the iterating.But in application simply doing like this will lead to a rough reconstructed signal.In this paper,the concepts of distinct coefficient COF and threshold coefficient TH are proposed to overcome the deficiency,and the step of edge extraction is also changed to make the choice of COF and TH conveniently.The edges of signal can be kept well while the noise is removed greatly by appropriately choosing COF and TH.The simulation shows that the filtering efficiency of the new algorithm is better than the original one.