Iranian Journal of Oil & Gas Science and Technology, Vol. x (xxxx), No. x, pp.

xx-xx
http://ijogst.put.ac.ir

Seismic Temporal Resolution Improvement Using

Logarithmic of Time-Frequency Transform Method
Saman Gholtashi1, and Amin Roshandel Kahoo1*
1

School of Mining, Petroleum and Geophysics Engineering, University of Shahrood, Shahrood,

Iran

Abstract
Improvement of the temporal resolution of seismic data is a critical issue in
hydrocarbon exploration. It is important for getting more detailed structural and
stratigraphic information. Many methods have been introduced to improve the
vertical resolution of reflection seismic data. Each of them has advantages and
disadvantages which are due to the assumptions and theories governing their issue.
In this paper, we improve the temporal resolution of reflection seismic data using
the logarithmic of time-frequency transform method. This method has minimum of
user-defined parameters. The algorithm uses valuable properties of both the timefrequency transform and the cepstrum to extending the frequency band at each
translation of the spectral decomposing window. In this method, displacement of
amplitude spectrum by its logarithm is the basic idea of the algorithm. We test the
mentioned algorithm on both synthetic and real data. The result of both tests shows
the introduced method can increase the temporal resolution of seismic data.
Keywords: Seismic Temporal Resolution, Time-Frequency Transform, Logarithmic
Method, Enhancing Temporal Resolution

1. Introduction
There are two types of resolution in surface reflection seismic data: horizontal
resolution and vertical resolution. Vertical or temporal resolution is expressed by the
tuning thickness and horizontal or spatial resolution is expressed by the Fresnel zone
(Badley, 1985). Improvement of the temporal resolution of seismic data is a critical
subject in hydrocarbon exploration and characterizing thin-layered hydrocarbon
reservoirs. It is getting for obtaining more detailed structural and stratigraphic
information.
Tuning thickness is defined as a quarter of the dominant wavelength at the position
of the target layer (Sheriff and Geldart, 1995). It is the minimum thickness, which

S. Gholtashi and A. Roshandel Kahoo / Seismic Temporal Resolution Enhancement Using 3

the top and bottom of a layer is separable. The tuning thickness is related to interval
velocity of target layer and dominant frequency of traveling wave at the depth of
the target layer. So, the increase in the dominant frequency of seismic data can help
to improve the temporal resolution.
Many methods have been introduced to increase the vertical resolution of reflection
seismic data. Inverse Q-filter (Wang, 2008) and different deconvolution methods
(Yilmaz and Doherty, 2001) are the basic methods of the resolution improvement. In
the deconvolution procedure, the band limited seismic source signature is
compressed by various types of method to increase the frequency band of seismic
source wavelet.
The wavelet transform and time-frequency representation are the basis of many
methods of vertical resolution improvement of seismic data (Shang and Caldwell,
2003; Matos and Marfurt, 2014; Sajid and Ghosh, 2014; Zhou et al., 2014).
In this paper, we used the cepstrum theory (Oppenheim et al., 1997) to improve the
temporal resolution of reflection seismic data. The inverse Fourier transform of the
logarithm of the amplitude spectrum of a signal is named the cepstrum (Sajid and
Ghosh, 2014). We extend the cepstrum to time-frequency representation. The
mentioned algorithm is applied to synthetic and real seismic data.

2. Methodology
First, we introduce the method of resolution improving based on cepstrum in Fourier
domain. This method consists of three steps. In the first step, the time domain

x t
f
signal,
, is transformed to frequency domain
by Fourier transform formula
(Proakis and Manolakis, 2007) as eq. (1). Then the amplitude and phase spectrum
calculated from the Fourier transform of signal as eq. (2).

(1)

X f

x t e jt dt

Amplitude Spectrum A f X f

(2)

Phase Spectrum f R X f

In the next step, we extend the amplitude spectrum of signal by replacing it with its
logarithm. The phase spectrum of the signal remains unchanged in this procedure.
In the third step, the minimum value of logarithmic amplitude spectrum subtracted
from all spectral values to make positive the logarithmic amplitude spectrum. In
order to make equal the total energy of logarithmic amplitude spectrum before and
after the whitening, the logarithmic amplitude spectrum is normalized by its total
energy. The high resolution seismic trace can be reconstructed by using the
normalized logarithmic amplitude spectrum and the unchanged phase spectrum.

S. Gholtashi and A. Roshandel Kahoo / Seismic Temporal Resolution Enhancement Using 3

We tested the mentioned algorithm on a 15 Hz Ricker wavelet (Sheriff and Geldart,
1995) and showed the results in Figure 1. Figure 1 (a, b, c) show the time domain
original Ricker wavelet and its amplitude and phase spectrum, respectively. The
normalized logarithmic amplitude spectrum and the original phase spectrum are
shown in Figure 1 (e, f). We calculated the inverse Fourier transform of the modified
amplitude and the original phase spectra to gain the improved resolution seismic
trace which is shown in Figure 1 (d). Comparing Figures 1 (a) and (d) showed that
the propose algorithm can be improved the temporal resolution of seismic trace.
Because of the Fourier transform's limitations in analyzing non-stationary signals such as seismic trace,
we replace the Fourier transform with time-frequency transform in above mentioned
algorithm. There are various types of time-frequency transforms (Boashash, 2003), but we used the
Short Time Fourier Transform (STFT) which is the usual and common time-frequency transform. STFT of
a time domain signal,

x t

, can be calculated as Eq. (3):

X t, f

(3)

x g t e

i 2 f

g t

X t, f

where
is a Gaussian window. In general, the
be represented by amplitude and phase as Eq. (4):

is a complex value and can

real X t , f imag X t , f
imag X t , f
t , f tan

real X t , f
A t, f

(4)

In Figure (2), we showed the time domain Ricker wavelet and its amplitude spectrum
of both Fourier and STFT transform.
The method of seismic resolution improving based on cepstrum in STFT domain also
consists of three steps (Sajid and Ghosh, 2014). In the first step, the trace
spectrogram is calculated through the STFT by using the Gaussian window as
described in Eq. (3). The second step is the calculation of the logarithmic amplitude
of the trace spectrogram as Eq. (5):
(5)

LF t , f log10 A t , f

Like the method based on Fourier transform, the logarithmic amplitude spectrum is
made purely positive by subtracting its minimum value from all spectral values. In
the final step, the normalized logarithmic amplitude of the trace spectrogram is
calculated as Eq. (6) through (7):
(6)

LFP t , f LF t , f min LF t , f

(7)

Iranian Journal of Oil & Gas Science and Technology, Vol. x (xxxx), No. x, pp. xx-xx

A t , f df
LFPE t , f LFP t , f
LFP t , f df

Now, we can obtain the high resolution seismic trace by using the modified
amplitude spectrum and the original phase spectrum as described in Eq. (8) through
Eq. (9):
(8)

(9)

X t , f LFPE t , f exp j t , f
x t

X , f g t ei 2 ft dfd

The method of seismic resolution improving based on cepstrum in STFT domain was
tested on 40 Hz Ricker wavelet. In Figure 3, we show the 40 Hz Ricker wavelet
before and after the application of the algorithm and their STFT spectrogram. As can
be seen, the proposed algorithm compacted the wavelet in time domain and
extended in frequency domain without generating false features.

3. Results
3.1. Synthetic data
To investigate the efficiency of the above mentioned algorithm, we test the proposed
method on a synthetic 2D seismic section. Figure 4 (a, b) shows the free-noise
synthetic seismic section and its average amplitude spectrum, respectively. The
synthetic data consist of a single reflector, a thin layer and a wedge-shape model.
Due to the low temporal resolution in original synthetic seismic section, we are not
able to separate the reflections from the top and bottom of the thin layer. Also, the
reflections from top and bottom of the edge are detectable at trace 25.
Result of applying the method on the synthetic seismic section is shown in Figure 4
(c). As can be seen, the temporal resolution of seismic section is dramatically
increased. Reflections from the top and bottom of thin layer are fully recognizable. It
can be observed in the wedge model that the reflections which started to full
interference at trace 25 were still separable up to trace 39. The average amplitude
spectrum of output seismic section is showed in Figure 4 (d).
To evaluate the performance of the method in the presence of noise, we added the
5% white Gaussian noise to the synthetic seismic section and apply the method on
noisy data. The results are shown in Figure 5. As can be seen, the algorithm
performance is good even in the presence of noise.

3.2. Field data

S. Gholtashi and A. Roshandel Kahoo / Seismic Temporal Resolution Enhancement Using 3

We applied the method on a field seismic data from one of hydrocarbon fields in
southwestern of Iran. The real seismic section before and after the application of
method are shown in Figure 6. It can be easily observed that the resolution of the
real seismic data was considerably increased, and many hidden features were
discovered. For a closer look, we chose three windows of data before and after
applying the algorithm and they were shown magnified in Figure 7. When comparing
their amplitude spectra as shown in Figure 8, it can be seen that the field seismic
data after the applying the algorithm have a broader amplitude spectrum than the
original real data.

4. Conclusion
We have introduced a new algorithm that improves the temporal resolution of
seismic data by using the logarithmic of time-frequency transform method. The
algorithm uses valuable properties of both the time-frequency transform and the
cepstrum to extending the frequency band at each translation of the spectral
decomposing window. The result of the application of the algorithm on both
synthetic and real seismic data show that the introduced method can increase the
temporal resolution of seismic data. Also, the results of the algorithm in the
presence of noise, show that the algorithm can improve the temporal resolution of
the seismic data without greatly boosting noise.

References
Badley, M. E., 1985, Practical seismic interpretation: International Human Resources
Development Corporation.
Boashash, B., 2003, Time Frequency Analysis: Elsevier Science.
Matos, M. C. d., and K. Marfurt, 2014, Complex wavelet transform spectral
broadening, SEG Technical Program Expanded Abstracts 2014, 1465-1469.
Oppenheim, A. V., A. S. Willsky, and S. H. Nawab, 1997, Signals and Systems:
Prentice Hall.
Proakis, J. G., and D. G. Manolakis, 2007, Digital Signal Processing, Principles,
Algorithms, and Applications: Pearson Prentice Hall.
Sajid, M., and D. Ghosh, 2014, Logarithm of short-time Fourier transform for
extending the seismic bandwidth: Geophysical Prospecting, 62, 11001110.
Shang, B. Z., and D. H. Caldwell, 2003, A bandwidth enhancement workflow through
wavelet analysis, SEG Technical Program Expanded Abstracts 2003, 2012-2015.
Sheriff, R. E., and L. P. Geldart, 1995, Exploration Seismology: Cambridge University
Press.
Wang, Y., 2008, Seismic Inverse Q Filtering: Wiley.
Yilmaz, ., and S. M. Doherty, 2001, Seismic Data Analysis: processing, inversion,
and interpretation of seismic data: Society of Exploration Geophysicists.
Zhou, H., C. Wang, K. J. Marfurt, Y. Jiang, and J. Bi, 2014, Enhancing the resolution of
seismic data using improved time-frequency spectral modeling, SEG Technical
Program Expanded Abstracts 2014, 2656-2661.

Iranian Journal of Oil & Gas Science and Technology, Vol. x (xxxx), No. x, pp. xx-xx

Figure 1

S. Gholtashi and A. Roshandel Kahoo / Seismic Temporal Resolution Enhancement Using 3

(a) Input wavelet: 15 Hz Ricer wavelet in time domain and its (b) amplitude and (c) phase spectrum. (d)
Reconstructed high resolution wavelet from the (e, f) the normalized logarithmic amplitude spectrum and the original
phase spectrum of input wavelet.

Figure 2
(a) 40 Hz Ricer wavelet in time domain and its spectrum of (b) Fourier transform and (c) short time Fourier
transform.

Iranian Journal of Oil & Gas Science and Technology, Vol. x (xxxx), No. x, pp. xx-xx

Figure 3
(a) Input wavelet: 40 Hz Ricer wavelet in time domain and (b) its STFT spectrogram. (c) Reconstructed wavelet and
(d) its STFT spectrogram.

S. Gholtashi and A. Roshandel Kahoo / Seismic Temporal Resolution Enhancement Using 3

Figure 4
(a) Free noise synthetic seismic section and (b) its average amplitude spectrum. (c) Synthetic seismic section after
the application of the propose method and (d) its average amplitude spectrum.

Iranian Journal of Oil & Gas Science and Technology, Vol. x (xxxx), No. x, pp. xx-xx

Figure 5
(a) Noisy synthetic seismic section and (b) its average amplitude spectrum. (c) Synthetic seismic section after the
application of the propose method and (d) its average amplitude spectrum.

S. Gholtashi and A. Roshandel Kahoo / Seismic Temporal Resolution Enhancement Using 3

Figure 6
Real seismic data (a) before and (b) after the application of the propose algorithm.

Iranian Journal of Oil & Gas Science and Technology, Vol. x (xxxx), No. x, pp. xx-xx

Figure 7
(a, b) The blue window in Figure 6 before and after the application of the propose algorithm. (c, d) The green
window in Figure 6 before and after the application of the propose algorithm. (e, f) The black window in Figure 6
before and after the application of the propose algorithm.

S. Gholtashi and A. Roshandel Kahoo / Seismic Temporal Resolution Enhancement Using 3

Figure 8
The average amplitude spectrum of real seismic section (a) before and (b) after the application of the propose
algorithm.