Amplitude preservation in multicomponent processing using local similarity
Khemraj Shukla, Priyank Jaiswal · 2015
Summary Frequency–space (f-x) deconvolution is a common practice in multicomponent denoising. While it helps in removing coherent and incoherent noise, requirement of different filter lengths in different components modulates amplitudes of the useful signals inconsistently. This present a problem in data rotation, which requires relative amplitudes in all components to remain unchanged through prior processing algorithms such as f-x deconvolution. Here we show that in the useful seismic energy that is attenuated in f-x deconvolution, known as the leaked energy, can be retrieved using the local similarity attribute, which is a spatio-temporal function of cross-correlation coefficients. The basic assumption is that the signal and noise components that make up the seismic coda are mutually orthogonal, i.e., they have no similarity. The similarity attribute between the filtered and the noise coda obtained from f-x deconvolution is the image of the leaked energy. When this similarity map is normalized and windowed between a minimum and maximum threshold, it acts like a weight function which can be used to extract leaked energy from the noise coda. In a multicomponent synthetic test when the leaked energy is restored using common minimum windowing threshold for all components, excellent preservation of relative amplitude is seen. Ongoing research tests this process with real multicomponent data from Rock Springs Uplift, Wyoming, with encouraging initial results.