模型约束分频局域去噪方法及应用

由于受地表条件和地质条件的影响,复杂地区地震资料的各种干扰波都十分发育并严重影响了地震资料的成像。目前最好的去噪方式是对不同的干扰波采取不同的去噪方法,去噪最彻底的方法是减去法。在常规去除规则干扰的处理中,常常在全频域采用带通滤波的方法,而有效波与噪音在频带上是相互重叠的,所以在去除噪音的同时不可避免地影响有效信号,导致低频成分的丧失。采用局域分频思路,将地震信号分成若干个频段,仅对噪音富集的频段剖面进行模型拾取减法去噪,既较干净地去除了噪音,又保护了有效信号,从而改善原始信噪比。

MODEL-RESTRICTED FREQUENCY-DIVISION LOCAL NOISE ATTENUATION

Seismic data in complex area contains varieties of abundant disturbing waves that badly affect the imaging of the data. Now, the best way for noise attenuation is to introduce different noise attenuation methods to diverse disturbing wave, among which the subtraction has more thorough-paced result. Bandfilters are commonly used in full frequency domain to eliminate regular noise in conventional processing. But the frequency of the reflection waves and the noise are overlapped. So the reflection waves would be damaged inevitably. Such process will lead to the loss of the low frequency divisions. The method introduced in this paper introduces frequency division theory in data processing. It divides the seismic signal into several sections of different frequency divisions, and only processes the section that has affluence noise with model-restrained subtraction noise attenuation method. Thus not only can it wipe completely out the noise, but preserve the reflection waves, and sequentially improves the ratio of signalto-noise. The examples show that the method is practicable and effective.