基于小波变换的分频最小光滑滤波去噪

MINIMUM SMOOTH FILTER AND FREQUENCY-DIVIDER BASED ON WAVELET TRANSFROM

在地震勘探中,随机噪音是一种频带较宽,严重影响有效波信噪比的干扰波。一般的去噪方法有:f-x、w-x、f-x奇异值分解等,但这些都只考虑横向上随机噪音的不相关性,而在一维方向上研究甚少。小波变换是一种时频分析的方法,根据它的分频和局部分析能力,结合一维方向上去噪的光滑滤波,就能有效地消除随机干扰,且保证了有效波的中、高频成份,经过小波包重构,可恢复有效波信号。该方法在去噪方面是十分有效的。

In the seismic survey, random noises with a wide frequency band can strongly influence the ratio of signal to noise. The conventional methods for the noise elimination are: f-x domain prediction? wx domain prediction? SVD of the fx domain etc. These methods have, however, limitations that they can only take into account irrelevance in horizontal direction with little study in 1D direction.Wavelet packet is an analytical time-frequency method. Basing on the characters of frequency dividing and analysis in local place and combining the smooth filter of the noise elimination in 1D direction, they can effectively remove the random noise, while at the same times insure middle frequency and high frequency components in effective wave. Then, the signals can be recovered from the wavelet reconstitution. The examples shows that the method is effective for the random noise elimination.