Intermediate Frequency Seismic Data De-noising by Radial Trace Time-Frequency Peak Filtering

In seismic data processing, recognizing and tracking the seismic events from records with strong random noise is an especially crucial step for further interpretation and analysis. Time-frequency peak filtering(TFPF), a time-frequency based method, has achieved many good results in DaQing oil basin in China for its accuracy in non-stationary signal estimation without prior knowledge but low SNR. However, for intermediate dominant frequency(above 40Hz), TFPF would bring serious bias even when the time window τ decreases to 3. In another word, intermediate frequency preservation and the bias limitation of TFPF is a pair of contradiction, which cannot be solved by shortening τ. Taking both the unbiased TFPF condition and the adjacent channels' correlation into considered, this paper introduces the radial trace(RT)transform to obtain a frequency reduced input and thus decrease the TFPF bias. First decrease the intermediate frequencies along radial traces close to the reflection event direction. Then obtain a less-biased TFPF estimation with suitable τ. At last, recover the intermediate frequency with the inverse RT. Both synthetic and field data experiments prove its better performance in intermediate frequency seismic data de-noising. Fig1 extracts a single channel with dominant frequency of 45Hz from a 2D synthetic record. RT TFPF behaves better in random noise attenuation(e.g. solid circles), and meanwhile introduces less bias (e.g. dotted circle). RT TFPF attenuates more random noise between 0-30Hz and 70-120Hz, and preserves intermediate frequencies around 30-70Hz. To field data in Fig2, RT TFPF recovers some reflection events more clearly, and some events become more continuous and compact. A single channel before and after different TFPF. Different TFPF results of field data.