摘要
在小波变换模极大值去噪原理的基础上 ,通过整体变分法估计模极大值以外、非零值的小波系数 ,然后经过小波逆变换重构信号 .为求解这个极小化问题 ,采用了改进的次梯度方法 .数值计算结果表明 :我们提出的方法有效抑制了小波变换中固有的伪Gibbs现象 ,重构信号的边缘、脉冲位置都十分准确 ,信噪比也得到明显改善 .
A novel hybrid method is presented, which combines Wavelet Transform Modulus Maxima and Total Variation, to improve reconstruction by way of using the nonzero wavelet coefficients besides modulus maxima. How to determine the nonzero wavelet coefficients is a minimization problem, an exact penalty function approach is used. Then an improved sub-gradient algorithm is utilized. The numerical experiments show that pseudo-Gibbs phenomena are restrained, the edge and peak location of reconstructed signal are very exact, and SNR is distinctly improved.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2004年第10期1713-1715,共3页
Acta Electronica Sinica
关键词
小波
整体变分
模极大值
信号去噪
信号重构
次梯度
Algorithms
Optimization
Signal filtering and prediction
Signal to noise ratio
Wavelet transforms
