Seismic Data Reconstruction via Shearlet-Regularized Directional Inpainting

Size: px
Start display at page:

Download "Seismic Data Reconstruction via Shearlet-Regularized Directional Inpainting"

Transcription

1 Seismic Data Reconstruction via Shearlet-Regularized Directional Inpainting Sören Häuser and Jianwei Ma May 15, 2012 We propose a new method for seismic data reconstruction by directional weighted shearlet-regularized image inpainting. The main idea is the application of a directional and scale dependent thresholding (sdt) to penalize the vertical artifact edges. The alternating direction method of multipliers (ADMM) is used to solve the resulting sparsity-promoting l 1 -norm regularization. Numerical experiments show a very good performance of the proposed method. 1 Introduction In applied geophysics various methods were developed to investigate the subsurface. Due to its ability to produce images down to depths of thousands of meters with a good resolution, the seismic method has become the most commonly used geophysical method in the oil and gas industry. The simplest possible seismic measurement would be a 1D pint measurement with a single source (often referred to as a shot) and receiver, both located in the same place. The results could be displayed as a seismic trace, which is a graph of the signal amplitude against travel time. Much more useful is a 2D measurement, with sources and receivers positioned along a line on the surface. In practice, many receivers regularly spaced along the line are used to record the signal from a series of source points. To avoid information losses, the seismic data should be sampled according to the Shannon-Nyquist criterion. However, seismic data are often sparsely or incompletely sampled along the spatial coordinates, partly caused by surface obstacles, dead trace, no-permit areas and economic constraints. So the reconstruction and interpolation of seismic data is a very important problem in the seismic community. The quality of reconstruction will impact the subsequent seismic processing steps, e.g., denoising, AVO, imaging, etc. Many methods have been proposed for the reconstruction of seismic data. These methods are mainly divided into two categories: wave-equation methods and signal processing methods. The former methods use the simulation of wave propagation to reconstruct seismic data, so Department of Mathematics, University of Kaiserslautern, Kaiserslautern, Germany, haeuser@mathematik.uni-kl.de Institute of Applied Mathematics, Harbin Institute of Technology, Harbin, China, jma@hit.edu.cn 1

2 that they require subsurface information (e.g., velocity distribution in earth medium), see e.g. [21, 26]. Most methods in the signal-processing-based category use sparse transforms such as the Radon transform [13, 28], Fourier transform [22, 29, 17], and curvelet transform [11, 12, 27, 18]. In particular, Herrmann et al. [11, 12, 27] applied the sparsity-promoting l 1 -norm minimization of curvelet coefficients, that leads to much improved results for edge/lineupspreserving reconstruction. It is well known that the curvelet transform sparsely represents seismic data that mainly includes line-like features. Thanks to the theory of compressed sensing, the seismic data can be efficiently reconstructed after random down-sampling (or taking less measurements) using a sparsity-promoting l 1 -norm convex optimization approach (even the sampling ratio is much less than the Shannon-Nyquist sampling ratio). This indicates that one can use less measured data (with more missing traces) to reconstruct the unknown seismic data with high precision. It could reduce the cost of seismic data survey. The method in this paper also follows this line. Instead of the curvelet transform we use the shearlet transform introduced in [14] and described in its fully finite discrete form in [10] in the sparsity-promoting l 1 -norm term. To our best knowledge this is the first time shearlets are used in the context of seismic data. Another popular techniques is the use of prediction filters. For instance, the low-frequency nonaliased components or observed data are used to build antialiasing prediction-error filters, and then the filters are applied to interpolate high-frequency components or missing traces. The prediction-filter method can be implemented in t x domain, f x domain, even curvelet domain, e.g., [25, 20, 4, 15, 18, 18]. Recently, rank-reducing methods are used for the seismic data reconstruction. In [19], Oropeza and Sacchi reorganized the seismic data into a Hankel matrix, and then use multichannel singular spectrum analysis (MSSA) to solve the rank-reduction problem of the Hankel matrix. In [31], Yang, Ma and Osher applied this theory and advanced fast algorithm of matrix completion (MC) to seismic data reconstruction. Many existing fast algorithms of nuclear-norm minimization can be applied in this framework. Recently, the authors [30] extended the lowrank method to a joint l 1 -norm of curvelet coefficient and nuclear-norm based low-rank method for seismic data reconstruction. In this paper, our contribution is two-fold. We will describe the missing traces recovery as an image inpainting problem using shearlets in the l 1 -sparsity regularization term. The regularization parameter Λ is usually chosen only scale dependent (st). We propose the additional use of the directional information given by the shearlet coefficients (scale-directional, sdt). In particular we use different weights for the horizontal and for the vertical directions to reduce visible artifacts. This can be done efficiently using the simple data structure provided by the recently published Fast Finite Shearlet Transform (FFST) [10, 9]. The organization of the paper is as follows: In Section 2 we introduce the shearlet transform and the basic steps for discretization. For our computations we will use FFST, an implementation of a translation invariant finite discrete shearlet transform discretized on the full grid. This is different to other implementations, see, e.g., ShearLab [24]. In Section 3 we apply the convex inpainting model to the missing trace reconstruction problem and show how to incorporate the shearlet transform into this model. We explain how the alternating direction method of multipliers (ADMM) can be used to find the minimizer of the functional exploiting the 2

3 Parseval-frame property of the discrete shearlet transform. In the last Section 4 we present the numerical results for two different data sets. 2 Shearlets and Shearlet transform In this first section we sketch the mathematical background of the shearlet transform that will be used as part of our inpainting algorithm. Similar to curvelets [2] shearlets were developed to provide a directional multiscale decomposition of images. However, the rotation used for curvelets to obtain the directional selectivity is replaced by a shear matrix S s that reads ( ) 1 s S s =, s R. 0 1 Note that the shearing is well suited for the discrete setting compared to the rotation used for the curvelet transform. Additionally shearlets stem from a square-integrable group representation, the so called shearlet group [5]. Similar to curvelets a shearlet is oriented in time (or space) domain and supported in a rectangle of width 2 2j and length 2 j, i.e. it fulfills the parabolic scaling relation width length 2. The shearlet ψ j,k,m emerges by dilation j, shear k and translation m from the mother shearlet ψ as (( 2 2j k2 ψ j,k,m (x) = ψ j ) ) 0 2 j (x m) We define the mother shearlet ψ L 2 (R 2 ) in the Fourier domain by ( ) ˆψ(ω) = ˆψ 1 (ω 1 ) ˆψ ω2 2 ω 1 where ψ 1 is usually the Meyer wavelet and ˆψ 2 is a bump function, for details see [9]. The dilated, sheared and translated shearlets in the Fourier domain are ˆψ j,k,m (ω) = ˆψ ( 1 4 j ) ω 1 ˆψ2 (2 j ω ) 2 + k e 2πi ω,m. ω 1 Fig. 1 shows a dilated and sheared shearlet in Fourier domain (left) and in time domain (right). The shearlet transform SH(f) of a function f L 2 (R 2 ) for a certain scale, shear and translation is the inner product between the function f and the respective shearlet ψ j,k,m, i.e., SH(f)(j, k, m) := f, ψ j,k,m = f(x)ψ j,k,m (x)dx. R 2 To get a good directional selectivity both from the horizontal and vertical point of view shearlets on the cone were introduced in [14]. We define the horizontal and the vertical cone by C h := {(ω 1, ω 2 ) R 2 : ω 1 1 2, ω 2 < ω 1 }, C v := {(ω 1, ω 2 ) R 2 : ω 2 1 2, ω 2 > ω 1 }, 3

4 (a) Shearlet in Fourier domain for j = 1 and k = 1 (b) Same shearlet in time domain (zoomed) Figure 1: Shearlet in Fourier and time domain. respectively. In the following, we consider digital images as functions sampled on the grid 1 N I := 1 N {(m 1, m 2 ) : m i = 0,..., N 1, i = 1, 2}, where we assume quadratic images for simplicity. Further, let j 0 := 1 2 log 2 N be the number of considered scales. To obtain a finite discrete shearlet transform, we use as finite discrete scaling, shear and translation parameters j = 0,..., j 0 1, 2 j k 2 j, t m := m N, m I. and discretize the frequency plane on Ω := { (ω 1, ω 2 ) : ω i = N 2,..., N } 2 1, i = 1, 2. We define the finite discrete shearlet transform as f, φ m for κ = 0, SH(f)(κ, j, k, m) = c(κ, j, k, m) := f, ψj,k,m κ for κ {h, v}, (1) f, ψ h v j,k,m for κ =, k = 2j. where j = 0,..., j 0 1, 2 j + 1 k 2 j 1 and m I if not stated in another way. The shearlet transform can be efficiently realized by applying the fft2 and its inverse ifft2. In view of the inverse shearlet transform it was proven in [9] that these discrete shearlets constitute a Parseval frame of the finite Euclidean space L 2 (I). Having a Parseval frame the inverse shearlet transform reads f = j j 1 κ {h,v} j=0 k= 2 j +1 m I j 0 1 f, ψj,k,m κ ψκ j,k,m + j=0 k=±2 j m I f, ψ h v j,k,m ψh v j,k,m + m I f, φ m φ m. (2) By (1) and (2) we can write the shearlet transform and its inverse in matrix-vector form as c = Sf and f = S c (3) where S S = I but SS I. The actual computation of f from given coefficients c(κ, j, k, m) := f, ψj,k,m κ is done in the Fourier domain. For details on the discretization and on computation we refer to [9]. 4

5 3 Inpainting Let the seismic measurement data be given as an image f R N N. We assume that the columns represent the spatial positions of the receivers and each row is a measurement at a certain time. The measured data are corrupted, i.e., a certain amount of columns is missing. The task is to recover the missing data. This can be seen as a general inpainting problem: Given an image with (some) regions of degraded pixels one wants to find a good approximation of the missing pixels. For a simple matrix-vector notation we assume the image to be column-wise reshaped as a vector. Thus, having an image f R N N we obtain the vectorized image vec(f) R N 2 where we retain the notation f for both the original and the reshaped image. Let C := {i {1,..., N 2 } : f(i) 0} be the set of indices of nonzero elements of f. With C we can define the mask H R N 2 N 2 as the diagonal matrix with H(i, i) = 1 if and only if i C and 0 otherwise. Hf is the application of the mask H to the image f. We want to find the recovered signal u as the minimizer of the functional { } 1 u := argmin 2 Hu f ΛSu 1 u R N2 where S is the shearlet transform applied to the vectorized image u in the sense of (3) and Λ is a diagonal matrix of regularization parameters. In our algorithm we will benefit from the fact that S S = I. (4) Alternating Direction Method of Multipliers To solve the problem in (4) we want to use the Alternating Direction Method of Multipliers (ADMM) see [1, 7, 8, 23]. To apply this algorithm we rewrite the problem (4) as { } 1 argmin 2 Hu f Λv 1 u R N2,v R ηn2 subject to Su = v (5) where η is the number of scales and shears in the shearlet transform S. Then the ADMM algorithm reads as follows: ADMM Algorithm for (5): Initialization: v (0), b (0) R ηn 2 and γ > 0. For i = 0,... repeat until a convergence criterion is reached: { 1 u (i+1) = argmin 2 Hu f } 2γ b(i) + Su v (i) 2 2, (6) u R N2 { v (i+1) = argmin Λv } 2γ b(i) + Su (i+1) v 2 2, (7) v R ηn2 b (i+1) = b (i) + Su (i+1) v (i+1). (8) The last step (8) is a simple update of b. We have to comment the first two steps. Due to the differentiability of (6) its solution follows by setting the gradient of the functional on the right-hand side to zero. Thus, the minimizer is given by the solution of 0 = H T (Hu f) + 1 ( γ S T Su + b (i) v (i)) 5

6 Since H T H = H, H T f = Hf = f and S T S = I N 2 no linear system has to be solved and we get u (i+1) = (H + 1γ ) 1 I N (f 2 1γ (b S T (i) v (i))) which means that we have just to apply an inverse shearlet transform. The matrix (H + 1 γ I N 2) is a diagonal matrix such that the inverse is the reciprocal of the diagonal elements. The second step (7) can be minimized as follows { v (i+1) = argmin Λv v R ηn2 = argmin v R ηn2 2γ v (b(i) + Su (i+1) ) }{{} { γ Λv v g 2 2 } =:g = S γλ (g) (componentwise soft-shrinkage). 2 2 } The soft-shrinkage S λ with a threshold λ is defined as x λ for x > λ, S λ (x) := 0 for x [ λ, λ], x + λ for x < λ. 3.1 The choice of Λ The regularization parameter Λ is a diagonal matrix in R ηn 2. Recall that η is the number of scales and shears of the shearlet transform and that we are thresholding the shearlet coefficients. The diagonal consists of η blocks of size N 2. The values within each block should be constant, otherwise we would threshold different regions of the image differently. Consequently, we characterize Λ by η variables. Further, these η variables represent the scale and shear, i.e., directional, information contained in the shearlet coefficients. We write Λ = (λ 0, λ 1,1,..., λ 1,4, λ 2,1,..., λ 2,8,..., λ j0,1,..., λ }{{}}{{} j0,2 j 0 +1 ) }{{} coarsest scale second scale finest scale where λ j,k is the parameter for the scale j and the direction k and λ 0 is the parameter for the low-pass which is usually set to zero. The idea behind our approach is now two-fold. As usual we use the shearlet transform to enforce the solution to be sparse in the shearlet domain, i.e., to contain curved structures, especially lines. This is reasonable since seismic data consists mostly of these kind of structures. For this approach it would be sufficient to choose the parameters in Λ only scale dependent, i.e, Λ = (λ 0, λ 1,..., λ j0 ). We will refer to this approach as only scale based thresholding (st). But we want to go one step further. The shearlet transform also contains directional information. Due to the missing columns a lot of additional (but unwanted) vertical edges are contained in the corrupted image. The original image in contrast has almost no edges in vertical direction. This is why we propose not to use the same parameter λ for all directions within 6

7 one scale but choose larger parameters for the vertical-like directions and smaller parameters for the horizontal-like directions. This results in penalizing vertical edges. We call this method scale-directional thresholding (sdt). But let us first illustrate this idea in more detail. In Fig. 2 we show the original data (left) and the corrupted data (right) with 50% of the columns missing. (a) original data X 0 (b) corrupted data with 50% of the columns missing Figure 2: Original data X 0 and corrupted data with 50% columns missing. For the shearlet coefficients of both images we plot in Fig. 3(a) for every scale and shear the mean absolute value of the shearlet coefficients over all translations, i.e, for given coefficients c we compute for scale j and respective shear k the value m j,k = 1 N 2 N x=1 y=1 N c(j, k, (x, y)). The regions between the vertical dashed lines represent the different scales and the annotation of the horizontal axis illustrates the respective orientation of the shearlet. The solid line is the mean value of the original data and the dashed line is the mean value of the corrupted data. Note that for a better comparison the mean value of the corrupted data is doubled. Comparing both graphs the different directions contained in the data become visible. The mean value of the corrupted data shows a peak at the most left position within each scale, i.e. the vertical direction. Clearly, the corrupted data contain much more large shearlet coefficients in vertical-like directions than the original data. For the horizontal-like directions the shearlet coefficients of the original and the corrupted data are similar, i.e. both data contain the same information. The difference between both lines is shown in Fig. 3(b). Again, we see the peaks for the vertical direction, i.e., large differences, and small values for the horizontal directions, i.e., small differences. Our regularization parameter Λ is set to these differences. During the numerical experiments it turned out that we obtain the best results when multiplying Λ with a factor α. This factor can be easily chosen to obtain the best SNR, i.e. we use αλ as regularization parameter. 7

8 Since the original image is not given in real applications we compute a reference image in the following way. In each row we (iteratively) replace every missing pixel with the mean value of its horizontal neighbors. In the first run these neighbors may be zero but with increasing iterations these gaps get filled from both sides. The SNR of the image reconstructed in this simple way is worse than our results presented below. But the difference of the shearlet coefficients compared to the ones of the original image is marginal such that we use this fast reconstructed image as reference image for computing Λ \ / \ / \ / \ / (a) mean absolute value of the shearlet coefficients for different scales and orientation, solid: original data, dashed: corrupted data (for a better visualization the mean value of the low-pass is set to zero) 0 0 \ / \ / \ / \ / (b) difference between mean absolute values of shearlet coefficients of corrupted and original data Figure 3: Comparison of the mean absolute values of the shearlet coefficients for the original and the corrupted data. 4 Numerical Results We want to present the results of our approach with two different data sets. The first one, a rather simple data set, has been shown in Fig. 2. We will refer to this data as X 0. As a second data set we choose a more complex one that is shown in Fig. 5(a) and the corrupted version in Fig 5(b). This data set is referred to as X 1. For both data sets we randomly set 50% of the columns to zero. As a performance measure we compute the signal-to-noise ratio (SNR) as ( X 2 ) F SNR = 10 log 10 X X 2 F where the original signal is denoted with X and the recovered signal with X. As a second measure we also compare the recovery of a the central trace, i.e., the central column of the matrix X. Therefore we plot the central trace of the original signal (solid line) and the same trace for the reconstructed signal (dashed line). In Fig. 4(a) we show the recovered signal X 0 and in 4(b) we compare the central traces of both the original and the recovered signal. The ADMM parameter γ only influences the convergence but not the result. For the given seismic data we obtained the best performance for γ = 8. 8

9 (a) Recovered data X 0 using scaledirectional thresholding, α = 1 6, SNR: (b) Comparison of central traces of the original signal (solid) and the recovered signal (dashed) Figure 4: Numerical results for dataset X 0 for scale-directional thresholding The recovered signal shows visually no difference to the original image and no artifacts are visible. A careful analysis of the trace comparison shows that the trace is recovered very well, especially for the larger amplitudes in the top. For the small amplitudes in the bottom some differences remain. For the more complex data set X 1 we obtain a similar result in Fig. 5. The first row of the figure shows the original signal on the left (Fig. 5(a)) and on the right the corrupted signal (Fig. 5(b)). The second row presents on left hand the recovered image (Fig. 5(c)) and on the right hand the trace comparison (Fig. 5(d)). Although, the SNR is worse than in the first experiment (due to the more complex structures), the recovered images has only very few remaining artifacts and no obvious difference is visible. The trace comparison also proves the very good reconstruction. For comparison we also present in Fig. 6 the results for the regularization parameters chosen only scale dependent (st), i.e., Λ = (λ 0, λ 1,..., λ j0 ). The parameters were again tuned for the best SNR. Each SNR is about 2-3 db worse compared to the first approach but additionally a lot of artifacts remain in the reconstructed images. At this point we should give a short heuristic explanation for these artifacts: As described above the main difference in the shearlet coefficients between the original and the corrupted image lies in the coefficients for the vertical directions. To reconstruct the image we just have to manipulate these coefficients. Thus we choose large parameters here and small parameters elsewhere. However, if we only have one parameter per scale we have to somehow find an average between these values. If the value is to small we end up with vertical artifacts, if the value is to big we will have artifacts in all directions because we manipulated to many coefficients. 9

10 (a) original data X 1 (b) corrupted data with 50% of the columns missing (c) Recovered data X 1 using scaledirectional thresholding, α = 1 8, SNR: (d) Comparison of central traces of the original signal (solid) and the recovered signal (dashed) Figure 5: Numerical results for dataset X 1 for scale-directional thresholding 10

11 Alternatively one could use for the images without noise an iterative algorithm which preserves the known pixels, see [3]. (a) Recovered data X 0 using only scale based thresholding for Λ = 1 (0, 0.1, 0.2, 0.4, 2), SNR: (b) Recovered data X 1 using only scale based thresholding for Λ = 1 (0, 0.1, 0.2, 0.4, 2), SNR: Figure 6: Numerical results for datasets X 0 and X 1 for only scale based thresholding To show the performance of our approach we now add white Gaussian noise to the data before applying the mask. Let σ denote the standard deviation. We start with a small σ = Since the noise will mostly effect the fine scales of the shearlet transform we adjust Λ by adding 0.1 to the values at the finest scale. Due to the small noise level the visual difference in the image is marginal and consequently we omit presenting the noisy images. Fig. 7 shows the results for both data sets X 0 and X 1, respectively. The second row shows the recovered images (X 0 in Fig. 7(c) and X 1 in Fig. 7(d)) using scale-directional weighting and the third row shows the trace comparisons to the respective original signal. In comparison to the non noisy data the SNR decreases slightly but visually the reconstruction is still very good for both data sets. However, the trace comparison shows more differences than before. The comparison with only scale-based weighting in the first row (X 0 in Fig. 7(a) and X 1 in Fig. 7(b)) shows that the results are still better for the additional directional weighting. For further comparisons we increase the noise level to σ = 0.1. As before we compare the directional thresholding with the scale based thresholding in Fig. 8. The first row shows the corrupted images, X 0 in Fig. 8(a) on the left and X 1 in Fig. 8(b) on the right. In the second row we show the results for the only scale based weighting. The last line shows the reconstruction with scale-directional weighting. We see that the the SNR is again smaller and a lot more artifacts appear for the only scale based weighting. Due to the higher noise the SNR decreases significantly in all cases but the main structures remain intact. We obtain small artifacts for the scale-directional weighting but much more for the only scale based weighting 11

12 (a) Recovered data X 0 with noise level σ = 0.01 and only scale based thresholding for Λ = 1 (0, 0.1, 0.2, 0.4, 2), SNR: (b) Recovered data X 1 with noise level σ = 0.01 and only scale based thresholding for Λ = 1 (0, 0.1, 0.2, 0.4, 2), SNR: (c) Recovered data X 0 with noise level σ = 0.01 and scale-directional thresholding for α = 1, SNR: (d) Recovered data X 1 with noise level σ = 0.01 and scale-directional thresholding for α = 1, SNR: (e) Comparison of central traces of the original signal X 0(solid) and the recovered signal (dashed) 12 (f) Comparison of central traces of the original signal X 1(solid) and the recovered signal (dashed) Figure 7: Numerical results for datasets X 0 and X 1 with noise σ = 0.01

13 noise level - σ = 0.01 σ = 0.1 data shearlets sdt st wavelets X X X X X X Table 1: Comparison of the SNR for different types of thresholds due to the thresholding over the complete fine scale. All parameters were chosen for the best SNR. Wavelet transform Using the wavelet transform in the regularizer and basic iterative thresholding [6, 16] with the iteration u (0) = 0, u (i+1) = W 1 S Λ W(u (i) + H T (f Hu (i) )) the results are worse than for the shearlet regularizer (in both cases). Obviously, shearlets (and curvelets) are better suited for these kind of curved structures. As wavelets we used Daubechies 6. Curvelet transform Instead of the shearlet transform it is also possible to use the curvelet transform in the ADMM. For the computation we used the publicly available toolbox Curve- Lab 1. The visual results and the SNR are very similar to the results presented here for the shearlet case, thus we omit presenting them here. It is also possible to implement the directional weighting in the curvelet setting. However, due to the more complex data structure (cell arrays) it is not as straight-forward as for the shearlet transform and the computation using CurveLab is about 2.5 times slower than using the FFST implementation of the shearlet transform. Summary In Table 1 we provide a summary of the SNR values for the different thresholds and regularizers. Comparing the SNR of both the scale-directional threshold (sdt) and the only scale based threshold (st) we see that the SNR is in all cases better for the scale-directional thresholding. The gab becomes smaller for higher noise level but the visual comparison of the results show that with the scale-directional thresholding we obtain much less artifacts. The performance of the wavelet regularizer is significant worse. This is why we omit listing the results for the noisy case here

14 (a) noisy and corrupted data X 0, σ = 0.1 (b) noisy and corrupted data X 1, σ = 0.1 (c) Recovered data X 0 with noise level σ = 0.1 and only scale based thresholding for Λ = 1 (0, 0.1, 0.2, 0.4, 1), SNR: (d) Recovered data X 1 with noise level σ = 0.1 and only scale based thresholding for Λ = 1 (0, 0.1, 0.2, 0.4, 1), SNR: (e) Recovered data X 0 with noise level σ = 0.1 and scale-directional thresholding for α = 1, SNR: (f) Recovered data X 1 with noise level σ = 0.1 and scale-directional thresholding for α = 1, SNR: Figure 8: Numerical results for datasets X 0 and X 1 with noise σ =

15 5 Conclusions In this paper we apply shearlet regularized image inpainting to recover corrupted seismic data. The provided results show the excellent performance of shearlet regularized inpainting by additionally using the directional information contained in the shearlet coefficients. Obviously this depends heavily on the structure of the missing elements. But seismic data often has strong directionality either through different oriented waves or because the different dimensions of the data represent different measured informations. References [1] S. Boyd, N. Parikh, E. Chu, B. Peleato, and J. Eckstein. Distributed optimization and statistical learning via the alternating direction method of multipliers. Foundations and Trends in Machine Learning, 3(1):1 122, [2] E. Candes, L. Demanet, D. Donoho, and L. Ying. Fast discrete curvelet transforms. Multiscale modeling and simulation, 5(3): , [3] R. Chan, S. Setzer, and G. Steidl. Inpainting by flexible haar-wavelet shrinkage. SIAM Journal on Imaging Sciences, 1(3): , [4] S. Crawley, R. Clapp, and J. Claerbout. Interpolation with smoothly nonstationary prediction-error filters. In 1999 SEG Annual Meeting, [5] S. Dahlke, G. Kutyniok, P. Maass, C. Sagiv, H.-G. Stark, and G. Teschke. The uncertainty principle associated with the continuous shearlet transform. International Journal on Wavelets Multiresolution and Information Processing, 6: , [6] I. Daubechies, M. Defrise, and C. De Mol. An iterative thresholding algorithm for linear inverse problems with a sparsity constraint. ArXiv Mathematics e-prints, July [7] J. Eckstein and D. P. Bertsekas. On the Douglas-Rachford splitting method and the proximal point algorithm for maximal monotone operators. Mathematical Programming, 55: , [8] D. Gabay. Applications of the method of multipliers to variational inequalities. In M. Fortin and R. Glowinski, editors, Augmented Lagrangian Methods: Applications to the Solution of Boundary Value Problems, chapter IX, pages North-Holland, Amsterdam, [9] S. Häuser. Fast finite shearlet transform: a tutorial. Preprint University of Kaiserslautern, [10] S. Häuser and G. Steidl. Convex multiclass segmentation with shearlet regularization. International Journal of Computer Mathematics, accepted, [11] G. Hennenfent and F. Herrmann. Simply denoise: wavefield reconstruction via jittered undersampling. Geophysics, 73(3):V19 V28,

16 [12] F. Herrmann and G. Hennenfent. Non-parametric seismic data recovery with curvelet frames. Geophysical Journal International, 173(1): , [13] M. Kabir and D. Verschuur. Restoration of missing offsets by parabolic radon transform. Geophysical Prospecting, 43(3): , [14] G. Kutyniok, K. Guo, and D. Labate. Sparse multidimensional representations using anisotropic dilation and shear operators. Wavelets und Splines (Athens, GA, 2005), G. Chen und MJ Lai, eds., Nashboro Press, Nashville, TN, pages , [15] Y. Liu and S. Fomel. Seismic data interpolation beyond aliasing using regularized nonstationary autoregression. Geophysics, 76:V69, [16] J. Ma and G. Plonka. The curvelet transform. IEEE Signal Processing Magazine, 27(2): , [17] M. Naghizadeh and K. Innanen. Seismic data interpolation using a fast generalized fourier transform. Geophysics, 76(1):V1 V10, [18] M. Naghizadeh and M. Sacchi. Beyond alias hierarchical scale curvelet interpolation of regularly and irregularly sampled seismic data. Geophysics, 75(6):WB189 WB202, [19] V. Oropeza and M. Sacchi. Simultaneous seismic data denoising and reconstruction via multichannel singular spectrum analysis. Geophysics, 76(3):V25 V32, [20] M. Porsani. Seismic trace interpolation using half-step prediction filters. Geophysics, 64:1461, [21] J. Ronen. Wave equation trace interpolation. Geophysics, 52(7): , [22] M. Sacchi, T. Ulrych, and C. Walker. Interpolation and extrapolation using a highresolution discrete fourier transform. Signal Processing, IEEE Transactions on, 46(1):31 38, [23] S. Setzer. Operator splittings, bregman methods and frame shrinkage in image processing. International Journal of Computer Vision, 92(3): , [24] M. Shahram, G. Kutyniok, and X. Zhuang. Shearlab: A rational design of a digital parabolic scaling algorithm. submitted. [25] S. Spitz. Seismic trace interpolation in the fx domain. Geophysics, 56(6): , [26] R. Stolt. Seismic data mapping and reconstruction. Geophysics, 67(3): , [27] G. Tang, R. Shahidi, J. Ma, and F. Herrmann. Applications of randomized sampling schemes to curvelet-based sparsity-promoting seismic data recovery,. Geophysical Prospecting, to appear. [28] D. Trad, T. Ulrych, and M. Sacchi. Accurate interpolation with high-resolution timevariant radon transforms. Geophysics, 67(2): , [29] S. Xu, Y. Zhang, D. Pham, and G. Lambaré. Antileakage fourier transform for seismic data regularization. Geophysics, 70(4):V87 V95, [30] Y. Yang and et al. et al. Joint l1- and nuclear-norm regularization for seismic data reconstruction. UCLA CAM Report,

17 [31] Y. Yang, J. Ma, and S. Osher. Seismic data reconstruction via matrix completion. UCLA CAM Report 12-14,

Seismic data interpolation using nonlinear shaping regularization a

Seismic data interpolation using nonlinear shaping regularization a Seismic data interpolation using nonlinear shaping regularization a a Published in Journal of Seismic Exploration, 24, no. 5, 327-342, (2015) Yangkang Chen, Lele Zhang and Le-wei Mo ABSTRACT Seismic data

More information

Mostafa Naghizadeh and Mauricio D. Sacchi

Mostafa Naghizadeh and Mauricio D. Sacchi Ground-roll elimination by scale and direction guided curvelet transform Mostafa Naghizadeh and Mauricio D. Sacchi Summary We propose a curvelet domain strategy to subtract ground-roll from seismic records.

More information

Seismic data interpolation beyond aliasing using regularized nonstationary autoregression a

Seismic data interpolation beyond aliasing using regularized nonstationary autoregression a Seismic data interpolation beyond aliasing using regularized nonstationary autoregression a a Published in Geophysics, 76, V69-V77, (2011) Yang Liu, Sergey Fomel ABSTRACT Seismic data are often inadequately

More information

G009 Scale and Direction-guided Interpolation of Aliased Seismic Data in the Curvelet Domain

G009 Scale and Direction-guided Interpolation of Aliased Seismic Data in the Curvelet Domain G009 Scale and Direction-guided Interpolation of Aliased Seismic Data in the Curvelet Domain M. Naghizadeh* (University of Alberta) & M. Sacchi (University of Alberta) SUMMARY We propose a robust interpolation

More information

A Nuclear Norm Minimization Algorithm with Application to Five Dimensional (5D) Seismic Data Recovery

A Nuclear Norm Minimization Algorithm with Application to Five Dimensional (5D) Seismic Data Recovery A Nuclear Norm Minimization Algorithm with Application to Five Dimensional (5D) Seismic Data Recovery Summary N. Kreimer, A. Stanton and M. D. Sacchi, University of Alberta, Edmonton, Canada kreimer@ualberta.ca

More information

Main Menu. Summary. sampled) f has a sparse representation transform domain S with. in certain. f S x, the relation becomes

Main Menu. Summary. sampled) f has a sparse representation transform domain S with. in certain. f S x, the relation becomes Preliminary study on Dreamlet based compressive sensing data recovery Ru-Shan Wu*, Yu Geng 1 and Lingling Ye, Modeling and Imaging Lab, Earth & Planetary Sciences/IGPP, University of California, Santa

More information

SUMMARY. In combination with compressive sensing, a successful reconstruction

SUMMARY. In combination with compressive sensing, a successful reconstruction Higher dimensional blue-noise sampling schemes for curvelet-based seismic data recovery Gang Tang, Tsinghua University & UBC-Seismic Laboratory for Imaging and Modeling (UBC-SLIM), Reza Shahidi, UBC-SLIM,

More information

Incoherent noise suppression with curvelet-domain sparsity Vishal Kumar, EOS-UBC and Felix J. Herrmann, EOS-UBC

Incoherent noise suppression with curvelet-domain sparsity Vishal Kumar, EOS-UBC and Felix J. Herrmann, EOS-UBC Incoherent noise suppression with curvelet-domain sparsity Vishal Kumar, EOS-UBC and Felix J. Herrmann, EOS-UBC SUMMARY The separation of signal and noise is a key issue in seismic data processing. By

More information

Hierarchical scale curvelet interpolation of aliased seismic data Mostafa Naghizadeh and Mauricio Sacchi

Hierarchical scale curvelet interpolation of aliased seismic data Mostafa Naghizadeh and Mauricio Sacchi Hierarchical scale curvelet interpolation of aliased seismic data Mostafa Naghizadeh and Mauricio Sacchi SUMMARY We propose a robust interpolation scheme for aliased regularly sampled seismic data that

More information

Randomized sampling strategies

Randomized sampling strategies Randomized sampling strategies Felix J. Herrmann SLIM Seismic Laboratory for Imaging and Modeling the University of British Columbia SLIM Drivers & Acquisition costs impediments Full-waveform inversion

More information

Geophysical Journal International

Geophysical Journal International Geophysical Journal International Geophys. J. Int. 215) 21, 118 1192 GJI Marine geosciences and applied geophysics doi: 1.193/gji/ggv72 Simultaneous seismic data interpolation and denoising with a new

More information

A low rank based seismic data interpolation via frequencypatches transform and low rank space projection

A low rank based seismic data interpolation via frequencypatches transform and low rank space projection A low rank based seismic data interpolation via frequencypatches transform and low rank space projection Zhengsheng Yao, Mike Galbraith and Randy Kolesar Schlumberger Summary We propose a new algorithm

More information

A least-squares shot-profile application of time-lapse inverse scattering theory

A least-squares shot-profile application of time-lapse inverse scattering theory A least-squares shot-profile application of time-lapse inverse scattering theory Mostafa Naghizadeh and Kris Innanen ABSTRACT The time-lapse imaging problem is addressed using least-squares shot-profile

More information

Seismic data reconstruction with Generative Adversarial Networks

Seismic data reconstruction with Generative Adversarial Networks Seismic data reconstruction with Generative Adversarial Networks Ali Siahkoohi 1, Rajiv Kumar 1,2 and Felix J. Herrmann 2 1 Seismic Laboratory for Imaging and Modeling (SLIM), The University of British

More information

Convex optimization algorithms for sparse and low-rank representations

Convex optimization algorithms for sparse and low-rank representations Convex optimization algorithms for sparse and low-rank representations Lieven Vandenberghe, Hsiao-Han Chao (UCLA) ECC 2013 Tutorial Session Sparse and low-rank representation methods in control, estimation,

More information

Journal of Applied Geophysics

Journal of Applied Geophysics Journal of Applied Geophysics 19 (214) 256 265 Contents lists available at ScienceDirect Journal of Applied Geophysics journal homepage: www.elsevier.com/locate/jappgeo Dreamlet-based interpolation using

More information

Beyond alias hierarchical scale curvelet interpolation of regularly and irregularly sampled seismic data

Beyond alias hierarchical scale curvelet interpolation of regularly and irregularly sampled seismic data Beyond alias hierarchical scale curvelet interpolation of regularly and irregularly sampled seismic data Mostafa Naghizadeh and Mauricio D. Sacchi University of Alberta, Department of Physics Edmonton,

More information

Curvelet Transform with Adaptive Tiling

Curvelet Transform with Adaptive Tiling Curvelet Transform with Adaptive Tiling Hasan Al-Marzouqi and Ghassan AlRegib School of Electrical and Computer Engineering Georgia Institute of Technology, Atlanta, GA, 30332-0250 {almarzouqi, alregib}@gatech.edu

More information

Two-dimensional fast generalized Fourier interpolation of seismic records

Two-dimensional fast generalized Fourier interpolation of seismic records 2D FGFT interpolation Two-dimensional fast generalized Fourier interpolation of seismic records Mostafa Naghizadeh and Kris Innanen ABSTRACT The fast generalized Fourier transform (FGFT) algorithm is extended

More information

A new sparse representation of seismic data using adaptive easy-path wavelet transform

A new sparse representation of seismic data using adaptive easy-path wavelet transform A new sparse representation of seismic data using adaptive easy-path wavelet transform Jianwei Ma 1,3, Gerlind Plonka 2, Hervé Chauris 3 1 Institute of Seismic Exploration, School of Aerospace, Tsinghua

More information

SEISMIC INTERPOLATION VIA CONJUGATE GRADIENT PURSUIT L. Fioretti 1, P. Mazzucchelli 1, N. Bienati 2 1

SEISMIC INTERPOLATION VIA CONJUGATE GRADIENT PURSUIT L. Fioretti 1, P. Mazzucchelli 1, N. Bienati 2 1 SEISMIC INTERPOLATION VIA CONJUGATE GRADIENT PURSUIT L. Fioretti 1, P. Mazzucchelli 1, N. Bienati 2 1 Aresys, Milano, Italy 2 eni E&P, San Donato Milanese, Italy Introduction. Seismic processing methods

More information

Multicomponent f-x seismic random noise attenuation via vector autoregressive operators

Multicomponent f-x seismic random noise attenuation via vector autoregressive operators Multicomponent f-x seismic random noise attenuation via vector autoregressive operators Mostafa Naghizadeh and Mauricio Sacchi ABSTRACT We propose an extension of the traditional frequency-space (f-x)

More information

Interpolation with pseudo-primaries: revisited

Interpolation with pseudo-primaries: revisited Stanford Exploration Project, Report 129, May 6, 2007, pages 85 94 Interpolation with pseudo-primaries: revisited William Curry ABSTRACT Large gaps may exist in marine data at near offsets. I generate

More information

Deconvolution with curvelet-domain sparsity Vishal Kumar, EOS-UBC and Felix J. Herrmann, EOS-UBC

Deconvolution with curvelet-domain sparsity Vishal Kumar, EOS-UBC and Felix J. Herrmann, EOS-UBC Deconvolution with curvelet-domain sparsity Vishal Kumar, EOS-UBC and Felix J. Herrmann, EOS-UBC SUMMARY We use the recently introduced multiscale and multidirectional curvelet transform to exploit the

More information

SUMMARY. denoise the original data at each iteration. This can be

SUMMARY. denoise the original data at each iteration. This can be A comparison of D reconstruction methods Aaron Stanton*, Nadia Kreimer, David Bonar, Mostafa Naghizadeh, and Mauricio Sacchi, Department of Physics, University of Alberta SUMMARY A comparison is made between

More information

Anisotropy-preserving 5D interpolation by hybrid Fourier transform

Anisotropy-preserving 5D interpolation by hybrid Fourier transform Anisotropy-preserving 5D interpolation by hybrid Fourier transform Juefu Wang and Shaowu Wang, CGG Summary We present an anisotropy-preserving interpolation method based on a hybrid 5D Fourier transform,

More information

Curvelet-domain multiple elimination with sparseness constraints. Felix J Herrmann (EOS-UBC) and Eric J. Verschuur (Delphi, TUD)

Curvelet-domain multiple elimination with sparseness constraints. Felix J Herrmann (EOS-UBC) and Eric J. Verschuur (Delphi, TUD) Curvelet-domain multiple elimination with sparseness constraints Felix J Herrmann (EOS-UBC) and Eric J. Verschuur (Delphi, TUD) Context SRME (Verschuur, Guitton, Berkhout, Weglein, Innanen) Sparse Radon

More information

3D pyramid interpolation

3D pyramid interpolation 3D pyramid interpolation Xukai Shen ABSTRACT Seismic data geometries are not always as nice and regular as we want due to various acquisition constraints. In such cases, data interpolation becomes necessary.

More information

Apex Shifted Radon Transform

Apex Shifted Radon Transform Apex Shifted Radon Transform Daniel Trad* Veritas DGC Inc., Calgary, Alberta, Canada dtrad@veritasdgc.com ABSTRACT The Apex Shifted Radon transform (ASRT) is an extension of the standard hyperbolic RT,

More information

SEG/New Orleans 2006 Annual Meeting

SEG/New Orleans 2006 Annual Meeting and its implications for the curvelet design Hervé Chauris, Ecole des Mines de Paris Summary This paper is a first attempt towards the migration of seismic data in the curvelet domain for heterogeneous

More information

A REPORT SUBMITTED TO THE DEPARTMENT OF ENERGY RESOURCES ENGINEERING OF STANFORD UNIVERSITY

A REPORT SUBMITTED TO THE DEPARTMENT OF ENERGY RESOURCES ENGINEERING OF STANFORD UNIVERSITY THE APPLICATION OF OPTSPACE ALGORITHM AND COMPARISON WITH LMAFIT ALGORITHM IN THREE- DIMENSIONAL SEISMIC DATA RECONSTRUCTION VIA LOW- RANK MATRIX COMPLETION A REPORT SUBMITTED TO THE DEPARTMENT OF ENERGY

More information

Five Dimensional Interpolation:exploring different Fourier operators

Five Dimensional Interpolation:exploring different Fourier operators Five Dimensional Interpolation:exploring different Fourier operators Daniel Trad CREWES-University of Calgary Summary Five-Dimensional interpolation has become a very popular method to pre-condition data

More information

Seismic data interpolation and de-noising in the frequency-wavenumber domain

Seismic data interpolation and de-noising in the frequency-wavenumber domain Seismic data interpolation and de-noising in the frequency-wavenumber domain Mostafa Naghizadeh ABSTRACT I introduce a unified approach for de-noising and interpolation of seismic data in the frequency-wavenumber

More information

Advanced phase retrieval: maximum likelihood technique with sparse regularization of phase and amplitude

Advanced phase retrieval: maximum likelihood technique with sparse regularization of phase and amplitude Advanced phase retrieval: maximum likelihood technique with sparse regularization of phase and amplitude A. Migukin *, V. atkovnik and J. Astola Department of Signal Processing, Tampere University of Technology,

More information

Seismic data Interpolation in the Continuous Wavenumber Domain, Flexibility and Accuracy

Seismic data Interpolation in the Continuous Wavenumber Domain, Flexibility and Accuracy Seismic data Interpolation in the Continuous Wavenumber Domain, Flexibility and Accuracy Ye Zheng Geo-X Exploration Services Inc. Summary Our recently developed algorithm, ASFT (Arbitrarily Sampled Fourier

More information

The Alternating Direction Method of Multipliers

The Alternating Direction Method of Multipliers The Alternating Direction Method of Multipliers With Adaptive Step Size Selection Peter Sutor, Jr. Project Advisor: Professor Tom Goldstein October 8, 2015 1 / 30 Introduction Presentation Outline 1 Convex

More information

An R Package flare for High Dimensional Linear Regression and Precision Matrix Estimation

An R Package flare for High Dimensional Linear Regression and Precision Matrix Estimation An R Package flare for High Dimensional Linear Regression and Precision Matrix Estimation Xingguo Li Tuo Zhao Xiaoming Yuan Han Liu Abstract This paper describes an R package named flare, which implements

More information

SUMMARY. These two projections are illustrated in the equation

SUMMARY. These two projections are illustrated in the equation Mitigating artifacts in Projection Onto Convex Sets interpolation Aaron Stanton*, University of Alberta, Mauricio D. Sacchi, University of Alberta, Ray Abma, BP, and Jaime A. Stein, Geotrace Technologies

More information

A dual domain algorithm for minimum nuclear norm deblending Jinkun Cheng and Mauricio D. Sacchi, University of Alberta

A dual domain algorithm for minimum nuclear norm deblending Jinkun Cheng and Mauricio D. Sacchi, University of Alberta A dual domain algorithm for minimum nuclear norm deblending Jinkun Cheng and Mauricio D. Sacchi, University of Alberta Downloaded /1/15 to 142.244.11.63. Redistribution subject to SEG license or copyright;

More information

Common-angle processing using reflection angle computed by kinematic pre-stack time demigration

Common-angle processing using reflection angle computed by kinematic pre-stack time demigration Common-angle processing using reflection angle computed by kinematic pre-stack time demigration Didier Lecerf*, Philippe Herrmann, Gilles Lambaré, Jean-Paul Tourré and Sylvian Legleut, CGGVeritas Summary

More information

Downloaded 09/01/14 to Redistribution subject to SEG license or copyright; see Terms of Use at

Downloaded 09/01/14 to Redistribution subject to SEG license or copyright; see Terms of Use at Random Noise Suppression Using Normalized Convolution Filter Fangyu Li*, Bo Zhang, Kurt J. Marfurt, The University of Oklahoma; Isaac Hall, Star Geophysics Inc.; Summary Random noise in seismic data hampers

More information

Detecting Burnscar from Hyperspectral Imagery via Sparse Representation with Low-Rank Interference

Detecting Burnscar from Hyperspectral Imagery via Sparse Representation with Low-Rank Interference Detecting Burnscar from Hyperspectral Imagery via Sparse Representation with Low-Rank Interference Minh Dao 1, Xiang Xiang 1, Bulent Ayhan 2, Chiman Kwan 2, Trac D. Tran 1 Johns Hopkins Univeristy, 3400

More information

Curvelet-domain matched filtering Felix J. Herrmann and Peyman P. Moghaddam EOS-UBC, and Deli Wang, Jilin University

Curvelet-domain matched filtering Felix J. Herrmann and Peyman P. Moghaddam EOS-UBC, and Deli Wang, Jilin University Curvelet-domain matched filtering Felix J. Herrmann and Peyman P. Moghaddam EOS-UBC, and Deli Wang, Jilin University SUMMARY Matching seismic wavefields and images lies at the heart of many pre-/post-processing

More information

Curvelet-based non-linear adaptive subtraction with sparseness constraints. Felix J Herrmann, Peyman P Moghaddam (EOS-UBC)

Curvelet-based non-linear adaptive subtraction with sparseness constraints. Felix J Herrmann, Peyman P Moghaddam (EOS-UBC) Curvelet-based non-linear adaptive subtraction with sparseness constraints Felix J Herrmann, Peyman P Moghaddam (EOS-UBC) Context Spiky/minimal structure deconvolution (Claerbout, Ulrych, Oldenburg, Sacchi,

More information

The flare Package for High Dimensional Linear Regression and Precision Matrix Estimation in R

The flare Package for High Dimensional Linear Regression and Precision Matrix Estimation in R Journal of Machine Learning Research 6 (205) 553-557 Submitted /2; Revised 3/4; Published 3/5 The flare Package for High Dimensional Linear Regression and Precision Matrix Estimation in R Xingguo Li Department

More information

Data interpolation in pyramid domain

Data interpolation in pyramid domain Data interpolation in pyramid domain Xukai Shen ABSTRACT Pyramid domain is defined as a frequency-space domain with different spatial grids for different frequencies. Data interpolation in pyramid domain

More information

Robust Seismic Image Amplitude Recovery Using Curvelets

Robust Seismic Image Amplitude Recovery Using Curvelets Robust Seismic Image Amplitude Recovery Using Curvelets Peyman P. Moghaddam Joint work with Felix J. Herrmann and Christiaan C. Stolk UNIVERSITY OF BRITISH COLUMBIA University of Twente References Curvelets

More information

Application of Proximal Algorithms to Three Dimensional Deconvolution Microscopy

Application of Proximal Algorithms to Three Dimensional Deconvolution Microscopy Application of Proximal Algorithms to Three Dimensional Deconvolution Microscopy Paroma Varma Stanford University paroma@stanford.edu Abstract In microscopy, shot noise dominates image formation, which

More information

SUMMARY. In this paper, we present a methodology to improve trace interpolation

SUMMARY. In this paper, we present a methodology to improve trace interpolation Reconstruction of seismic wavefields via low-rank matrix factorization in the hierarchical-separable matrix representation Rajiv Kumar, Hassan Mansour, Aleksandr Y. Aravkin, and Felix J. Herrmann University

More information

Sparse wavelet expansions for seismic tomography: Methods and algorithms

Sparse wavelet expansions for seismic tomography: Methods and algorithms Sparse wavelet expansions for seismic tomography: Methods and algorithms Ignace Loris Université Libre de Bruxelles International symposium on geophysical imaging with localized waves 24 28 July 2011 (Joint

More information

Writing Kirchhoff migration/modelling in a matrix form

Writing Kirchhoff migration/modelling in a matrix form Writing Kirchhoff migration/modelling in a matrix form Abdolnaser Yousefzadeh and John C. Bancroft Kirchhoff migration matrix ABSTRACT Kirchhoff prestack migration and modelling are linear operators. Therefore,

More information

Image denoising using curvelet transform: an approach for edge preservation

Image denoising using curvelet transform: an approach for edge preservation Journal of Scientific & Industrial Research Vol. 3469, January 00, pp. 34-38 J SCI IN RES VOL 69 JANUARY 00 Image denoising using curvelet transform: an approach for edge preservation Anil A Patil * and

More information

Robust Principal Component Analysis (RPCA)

Robust Principal Component Analysis (RPCA) Robust Principal Component Analysis (RPCA) & Matrix decomposition: into low-rank and sparse components Zhenfang Hu 2010.4.1 reference [1] Chandrasekharan, V., Sanghavi, S., Parillo, P., Wilsky, A.: Ranksparsity

More information

Amplitude and kinematic corrections of migrated images for non-unitary imaging operators

Amplitude and kinematic corrections of migrated images for non-unitary imaging operators Stanford Exploration Project, Report 113, July 8, 2003, pages 349 363 Amplitude and kinematic corrections of migrated images for non-unitary imaging operators Antoine Guitton 1 ABSTRACT Obtaining true-amplitude

More information

Practical implementation of SRME for land multiple attenuation

Practical implementation of SRME for land multiple attenuation Practical implementation of SRME for land multiple attenuation Juefu Wang* and Shaowu Wang, CGGVeritas, Calgary, Canada juefu.wang@cggveritas.com Summary We present a practical implementation of Surface

More information

Least squares Kirchhoff depth migration: important details

Least squares Kirchhoff depth migration: important details Least squares Kirchhoff depth migration: important details Daniel Trad CREWES-University of Calgary Summary Least squares migration has been an important research topic in the academia for about two decades,

More information

SUMMARY. Pursuit De-Noise (BPDN) problem, Chen et al., 2001; van den Berg and Friedlander, 2008):

SUMMARY. Pursuit De-Noise (BPDN) problem, Chen et al., 2001; van den Berg and Friedlander, 2008): Controlling linearization errors in l 1 regularized inversion by rerandomization Ning Tu, Xiang Li and Felix J. Herrmann, Earth and Ocean Sciences, University of British Columbia SUMMARY Linearized inversion

More information

Seismic wavefield inversion with curvelet-domain sparsity promotion Felix J. Herrmann, EOS-UBC and Deli Wang, Jilin University

Seismic wavefield inversion with curvelet-domain sparsity promotion Felix J. Herrmann, EOS-UBC and Deli Wang, Jilin University Seismic wavefield inversion with curvelet-domain sparsity promotion Felix J. Herrmann, EOS-UBC and Deli Wang, Jilin University SUMMARY Inverting seismic wavefields lies at the heart of seismic data processing

More information

SEG Houston 2009 International Exposition and Annual Meeting

SEG Houston 2009 International Exposition and Annual Meeting Yu Geng* 1, Ru-Shan Wu and Jinghuai Gao 2 Modeling and Imaging Laboratory, IGPP, University of California, Santa Cruz, CA 95064 Summary Local cosine/sine basis is a localized version of cosine/sine basis

More information

IMAGE DE-NOISING IN WAVELET DOMAIN

IMAGE DE-NOISING IN WAVELET DOMAIN IMAGE DE-NOISING IN WAVELET DOMAIN Aaditya Verma a, Shrey Agarwal a a Department of Civil Engineering, Indian Institute of Technology, Kanpur, India - (aaditya, ashrey)@iitk.ac.in KEY WORDS: Wavelets,

More information

Effects of multi-scale velocity heterogeneities on wave-equation migration Yong Ma and Paul Sava, Center for Wave Phenomena, Colorado School of Mines

Effects of multi-scale velocity heterogeneities on wave-equation migration Yong Ma and Paul Sava, Center for Wave Phenomena, Colorado School of Mines Effects of multi-scale velocity heterogeneities on wave-equation migration Yong Ma and Paul Sava, Center for Wave Phenomena, Colorado School of Mines SUMMARY Velocity models used for wavefield-based seismic

More information

Anisotropic representations for superresolution of hyperspectral data

Anisotropic representations for superresolution of hyperspectral data Anisotropic representations for superresolution of hyperspectral data Edward H. Bosch, Wojciech Czaja, James M. Murphy, and Daniel Weinberg Norbert Wiener Center Department of Mathematics University of

More information

Non-stationary interpolation in the f-x domain

Non-stationary interpolation in the f-x domain Stanford Exploration Project, Report 129, May 6, 2007, pages 75 85 Non-stationary interpolation in the f-x domain William Curry ABSTRACT Interpolation of seismic data has previously been performed using

More information

1D internal multiple prediction in a multidimensional world: errors and recommendations

1D internal multiple prediction in a multidimensional world: errors and recommendations 1D internal multiple prediction 1D internal multiple prediction in a multidimensional world: errors and recommendations Pan Pan and Kris Innanen ABSTRACT Internal multiples are more difficult to estimate

More information

Optimized Compressed Sensing for Curvelet-based. Seismic Data Reconstruction

Optimized Compressed Sensing for Curvelet-based. Seismic Data Reconstruction Optimized Compressed Sensing for Curvelet-based Seismic Data Reconstruction Wen Tang 1, Jianwei Ma 1, Felix J. Herrmann 2 1 Institute of Seismic Exploration, School of Aerospace, Tsinghua University, Beijing

More information

G009 Multi-dimensional Coherency Driven Denoising of Irregular Data

G009 Multi-dimensional Coherency Driven Denoising of Irregular Data G009 Multi-dimensional Coherency Driven Denoising of Irregular Data G. Poole* (CGGVeritas Services (UK) Ltd) SUMMARY Many land and ocean bottom datasets suffer from high levels of noise which make the

More information

Compressed Sensing for Electron Tomography

Compressed Sensing for Electron Tomography University of Maryland, College Park Department of Mathematics February 10, 2015 1/33 Outline I Introduction 1 Introduction 2 3 4 2/33 1 Introduction 2 3 4 3/33 Tomography Introduction Tomography - Producing

More information

Seismic data restoration via data-driven tight frame

Seismic data restoration via data-driven tight frame GEOPHYSICS, VOL. 79, NO. 3 (MAY-JUNE 214); P. V65 V74, 7 FIGS., 4 TABLES. 1.119/GEO213-252.1 Downloaded 4/2/14 to 129.116.232.214. Redistribution subject to SEG license or copyright; see Terms of Use at

More information

Th C 02 Model-Based Surface Wave Analysis and Attenuation

Th C 02 Model-Based Surface Wave Analysis and Attenuation Th C 02 Model-Based Surface Wave Analysis and Attenuation J. Bai* (Paradigm Geophysical), O. Yilmaz (Paradigm Geophysical) Summary Surface waves can significantly degrade overall data quality in seismic

More information

Main Menu. Summary. Introduction

Main Menu. Summary. Introduction Noise-thresholding sparse-spike inversion with global convergence: calibration and applications Douglas S. Sassen* and Michael Glinsky, Ion Geophysical Corp. Summary We demonstrate an innovative inversion

More information

Time-jittered ocean bottom seismic acquisition Haneet Wason and Felix J. Herrmann

Time-jittered ocean bottom seismic acquisition Haneet Wason and Felix J. Herrmann Time-jittered ocean bottom seismic acquisition Haneet Wason and Felix J. Herrmann SLIM University of British Columbia Challenges Need for full sampling - wave-equation based inversion (RTM & FWI) - SRME/EPSI

More information

A MAP Algorithm for AVO Seismic Inversion Based on the Mixed (L 2, non-l 2 ) Norms to Separate Primary and Multiple Signals in Slowness Space

A MAP Algorithm for AVO Seismic Inversion Based on the Mixed (L 2, non-l 2 ) Norms to Separate Primary and Multiple Signals in Slowness Space A MAP Algorithm for AVO Seismic Inversion Based on the Mixed (L 2, non-l 2 ) Norms to Separate Primary and Multiple Signals in Slowness Space Harold Ivan Angulo Bustos Rio Grande do Norte State University

More information

The Benefit of Tree Sparsity in Accelerated MRI

The Benefit of Tree Sparsity in Accelerated MRI The Benefit of Tree Sparsity in Accelerated MRI Chen Chen and Junzhou Huang Department of Computer Science and Engineering, The University of Texas at Arlington, TX, USA 76019 Abstract. The wavelet coefficients

More information

Ripplet: a New Transform for Feature Extraction and Image Representation

Ripplet: a New Transform for Feature Extraction and Image Representation Ripplet: a New Transform for Feature Extraction and Image Representation Dr. Dapeng Oliver Wu Joint work with Jun Xu Department of Electrical and Computer Engineering University of Florida Outline Motivation

More information

Bayesian Spherical Wavelet Shrinkage: Applications to Shape Analysis

Bayesian Spherical Wavelet Shrinkage: Applications to Shape Analysis Bayesian Spherical Wavelet Shrinkage: Applications to Shape Analysis Xavier Le Faucheur a, Brani Vidakovic b and Allen Tannenbaum a a School of Electrical and Computer Engineering, b Department of Biomedical

More information

Generalized Tree-Based Wavelet Transform and Applications to Patch-Based Image Processing

Generalized Tree-Based Wavelet Transform and Applications to Patch-Based Image Processing Generalized Tree-Based Wavelet Transform and * Michael Elad The Computer Science Department The Technion Israel Institute of technology Haifa 32000, Israel *Joint work with A Seminar in the Hebrew University

More information

SUMMARY METHOD. d(t 2 = τ 2 + x2

SUMMARY METHOD. d(t 2 = τ 2 + x2 Yujin Liu, Zhi Peng, William W. Symes and Wotao Yin, China University of Petroleum (Huadong), Rice University SUMMARY The Radon transform suffers from the typical problems of loss of resolution and aliasing

More information

Recent developments in curvelet-based seismic processing

Recent developments in curvelet-based seismic processing 0000 Recent developments in curvelet-based seismic processing Felix J. Herrmann (fherrmann@eos.ubc.ca) Seismic Laboratory for Imaging and Modeling Department of Earth and Ocean Sciences The University

More information

Stochastic conjugate gradient method for least-square seismic inversion problems Wei Huang*, Hua-Wei Zhou, University of Houston

Stochastic conjugate gradient method for least-square seismic inversion problems Wei Huang*, Hua-Wei Zhou, University of Houston Stochastic conjugate gradient method for least-square seismic inversion problems Wei Huang*, Hua-Wei Zhou, University of Houston Summary With the development of computational power, there has been an increased

More information

SUMMARY. min. m f (m) s.t. m C 1

SUMMARY. min. m f (m) s.t. m C 1 Regularizing waveform inversion by projections onto convex sets application to the D Chevron synthetic blind-test dataset Bas Peters *, Zhilong Fang *, Brendan Smithyman #, Felix J. Herrmann * * Seismic

More information

SUMMARY. forward/inverse discrete curvelet transform matrices (defined by the fast discrete curvelet transform, FDCT, with wrapping

SUMMARY. forward/inverse discrete curvelet transform matrices (defined by the fast discrete curvelet transform, FDCT, with wrapping Seismic data processing with curvelets: a multiscale and nonlinear approach Felix J. Herrmann, EOS-UBC, Deli Wang, Jilin University and Gilles Hennenfent and Peyman Moghaddam SUMMARY In this abstract,

More information

One-norm regularized inversion: learning from the Pareto curve

One-norm regularized inversion: learning from the Pareto curve One-norm regularized inversion: learning from the Pareto curve Gilles Hennenfent and Felix J. Herrmann, Earth & Ocean Sciences Dept., the University of British Columbia ABSTRACT Geophysical inverse problems

More information

A CURVELET-BASED DISTANCE MEASURE FOR SEISMIC IMAGES. Yazeed Alaudah and Ghassan AlRegib

A CURVELET-BASED DISTANCE MEASURE FOR SEISMIC IMAGES. Yazeed Alaudah and Ghassan AlRegib A CURVELET-BASED DISTANCE MEASURE FOR SEISMIC IMAGES Yazeed Alaudah and Ghassan AlRegib Center for Energy and Geo Processing (CeGP) at Georgia Tech and KFUPM School of Electrical and Computer Engineering

More information

RESTORING ARTIFACT-FREE MICROSCOPY IMAGE SEQUENCES. Robotics Institute Carnegie Mellon University 5000 Forbes Ave, Pittsburgh, PA 15213, USA

RESTORING ARTIFACT-FREE MICROSCOPY IMAGE SEQUENCES. Robotics Institute Carnegie Mellon University 5000 Forbes Ave, Pittsburgh, PA 15213, USA RESTORING ARTIFACT-FREE MICROSCOPY IMAGE SEQUENCES Zhaozheng Yin Takeo Kanade Robotics Institute Carnegie Mellon University 5000 Forbes Ave, Pittsburgh, PA 15213, USA ABSTRACT Phase contrast and differential

More information

Image Enhancement Techniques for Fingerprint Identification

Image Enhancement Techniques for Fingerprint Identification March 2013 1 Image Enhancement Techniques for Fingerprint Identification Pankaj Deshmukh, Siraj Pathan, Riyaz Pathan Abstract The aim of this paper is to propose a new method in fingerprint enhancement

More information

Introduction. Wavelets, Curvelets [4], Surfacelets [5].

Introduction. Wavelets, Curvelets [4], Surfacelets [5]. Introduction Signal reconstruction from the smallest possible Fourier measurements has been a key motivation in the compressed sensing (CS) research [1]. Accurate reconstruction from partial Fourier data

More information

Limitations of Matrix Completion via Trace Norm Minimization

Limitations of Matrix Completion via Trace Norm Minimization Limitations of Matrix Completion via Trace Norm Minimization ABSTRACT Xiaoxiao Shi Computer Science Department University of Illinois at Chicago xiaoxiao@cs.uic.edu In recent years, compressive sensing

More information

Multi-step auto-regressive reconstruction of seismic records

Multi-step auto-regressive reconstruction of seismic records Multi-step auto-regressive reconstruction of seismic records Mostafa Naghizadeh and Mauricio D. Sacchi ABSTRACT Linear prediction filters in the f-x domain are widely used to interpolate regularly sampled

More information

Short Note. Non-stationary PEFs and large gaps. William Curry 1 INTRODUCTION

Short Note. Non-stationary PEFs and large gaps. William Curry 1 INTRODUCTION Stanford Exploration Project, Report 120, May 3, 2005, pages 247 257 Short Note Non-stationary PEFs and large gaps William Curry 1 INTRODUCTION Prediction-error filters (PEFs) may be used to interpolate

More information

Reconstruction of irregularly sampled and aliased data with linear prediction filters

Reconstruction of irregularly sampled and aliased data with linear prediction filters Reconstruction o irregularly sampled and aliased data with linear prediction ilters Mostaa Naghizadeh and Mauricio Sacchi Signal Analysis and Imaging Group (SAIG) Department o Physics University o Alberta

More information

Interpolation using asymptote and apex shifted hyperbolic Radon transform

Interpolation using asymptote and apex shifted hyperbolic Radon transform Interpolation using asymptote and apex shifted hyperbolic Radon transform Amr Ibrahim, Paolo Terenghi and Mauricio D. Sacchi Department of Physics, University of Alberta PGS Abstract The asymptote and

More information

Sub-Nyquist sampling and sparsity: getting more information from fewer samples

Sub-Nyquist sampling and sparsity: getting more information from fewer samples Sub-Nyquist sampling and sparsity: getting more information from fewer samples Felix J. Herrmann SLIM Seismic Laboratory for Imaging and Modeling the University of British Columbia SLIM Drivers Our incessant

More information

Multi-azimuth velocity estimation

Multi-azimuth velocity estimation Stanford Exploration Project, Report 84, May 9, 2001, pages 1 87 Multi-azimuth velocity estimation Robert G. Clapp and Biondo Biondi 1 ABSTRACT It is well known that the inverse problem of estimating interval

More information

Full waveform inversion of physical model data Jian Cai*, Jie Zhang, University of Science and Technology of China (USTC)

Full waveform inversion of physical model data Jian Cai*, Jie Zhang, University of Science and Technology of China (USTC) of physical model data Jian Cai*, Jie Zhang, University of Science and Technology of China (USTC) Summary (FWI) is a promising technology for next generation model building. However, it faces with many

More information

Edge Detections Using Box Spline Tight Frames

Edge Detections Using Box Spline Tight Frames Edge Detections Using Box Spline Tight Frames Ming-Jun Lai 1) Abstract. We present a piece of numerical evidence that box spline tight frames are very useful for edge detection of images. Comparsion with

More information

Compressive Sensing for Multimedia. Communications in Wireless Sensor Networks

Compressive Sensing for Multimedia. Communications in Wireless Sensor Networks Compressive Sensing for Multimedia 1 Communications in Wireless Sensor Networks Wael Barakat & Rabih Saliba MDDSP Project Final Report Prof. Brian L. Evans May 9, 2008 Abstract Compressive Sensing is an

More information

Th ELI1 12 Joint Crossline Reconstruction and 3D Deghosting of Shallow Seismic Events from Multimeasurement Streamer Data

Th ELI1 12 Joint Crossline Reconstruction and 3D Deghosting of Shallow Seismic Events from Multimeasurement Streamer Data Th ELI1 12 Joint Crossline Reconstruction and 3D Deghosting of Shallow Seismic Events from Multimeasurement Streamer Data Y.I. Kamil* (Schlumberger), M. Vassallo (Schlumberger), W. Brouwer (Schlumberger),

More information

Image denoising in the wavelet domain using Improved Neigh-shrink

Image denoising in the wavelet domain using Improved Neigh-shrink Image denoising in the wavelet domain using Improved Neigh-shrink Rahim Kamran 1, Mehdi Nasri, Hossein Nezamabadi-pour 3, Saeid Saryazdi 4 1 Rahimkamran008@gmail.com nasri_me@yahoo.com 3 nezam@uk.ac.ir

More information

SUMMARY THEORY. L q Norm Reflectivity Inversion

SUMMARY THEORY. L q Norm Reflectivity Inversion Optimal L q Norm Regularization for Sparse Reflectivity Inversion Fangyu Li, University of Oklahoma & University of Georgia; Rui Xie, University of Georgia; WenZhan Song, University of Georgia & Intelligent

More information

P257 Transform-domain Sparsity Regularization in Reconstruction of Channelized Facies

P257 Transform-domain Sparsity Regularization in Reconstruction of Channelized Facies P257 Transform-domain Sparsity Regularization in Reconstruction of Channelized Facies. azemi* (University of Alberta) & H.R. Siahkoohi (University of Tehran) SUMMARY Petrophysical reservoir properties,

More information