f-k Filter and Deconvolution
Real data have noise of various kinds. Some signal enhancement programs are used prestack (before stacking) and some are used post-stack (after stacking CMPs). In the previous chapter, we used gain and band-pass filtering on prestack data. We only used one option out of many in sugain, but you are free to test other sugain options. In this chapter, we present two popular signal enhancement tools. These tools could prove useful in processing the Taiwan data set; however, we were unable to improve that data set using them. (Perhaps you can find a way!). To demonstrate their capability, we use them on Oz files (shot gathers) that we get from the Colorado School of Mines Center for Wave Phenomena web site (Section 3.3).
The first tool is a frequency-wavenumber (f-k) filter, a two-dimensional filter. We see how to use an f-k filter through the sudipfilt program. For a brief description of the f-k domain, see Sheriff (2002): f-k domain.
The second tool is deconvolution. We implement a deconvolution (decon) script that uses suacor, an autocorrelation program, and supef, a prediction-error filter.
In Section 10.4.1, we used the one-dimensional transform display script fxdisp. Script fxdisp transforms each trace from a time series to a frequency series. Here, we use interactive script ifk to apply a two-dimensional frequency filter. The f-k filter slopes can be specified in relative units of time over trace spacing (for example, seconds/meter) or in absolute units of time samples per trace. In this section, we specify the