用网络法提高地震资料信噪比

Improving the signal-to-noise ratio of seismic data by the network method

K－L变换用于地震记录去噪已收到了很好效果。但其计算量之大影响了K－L变换的实用价值。实际上，将K－L变换用于去噪时并不需要求解协方差阵的全部特征向量．只需要求解部分特征向量。本文用网络法实现K－L变换，它可以只求解部分特征向量．从而大大节省了运算时间．使K－L变换有可能在地震勘探中成为实用的去噪方法。此外．该网络算法具有良好的并行性．易于用VLSI阵列处理器实现。文中提供了实际资料处理的例子。

The K-L transform has been found effeetive when used to suppress noises of seismic data. But the large computation amount limnits its practical value. In fact, when using the K-L transform to Suppress noises. only part of characteristic vectors of the covariance matrix need to be solved instead of all the characteristic vectors. The paper proposes a network method to perform the K-L transform. which can find the necessary characteristic vectors one by one.Thus, the computation time is saved greatly, and the K-L transform promises to be a practical method in seismic cxploration.Moreover ,the method has a good parallelism in computation. and is easy to accomplished by VLSI processor array. An example for teal data processing is given.