Metta Innovations @ 15th CISBGf “Noise Attenuation by Anisotropic Diffusion Preserving Structural Features”


At the 15th CISBGf (International Congress Of Brazilian Geophysics Society) that happened between July 31 and August 3 in Rio de Janeiro our researcher engineer Patrícia Pampanelli presented the work “Noise Attenuation by Anisotropic Diffusion Preserving Structural Features” during the Workshop on Computational Geophysics. In this work we can see how HPC can be used on Geophysics domain. We also compare the computational performance of CPU and GPU implementations and show that GPU is about 14 times faster. The following abstract and the result images show us more detail of this work.


We introduce an edge-preserving filtering method based on the anisotropic diffusion of the amplitude field. Our formulation of the diffusion process uses seismic attributes to identify horizons and faults, which are used as constraints of the proposed method. Differently from previous works, we use the instantaneous phase gradient to orient the diffusion process, and an implicit method for solving the diffusion equation. For each filtering step, a system of equations must be resolved. The proposed method was applied to synthetic and real seismic data. The results show that this filter is effective in removing noise, while preserving structural features. Moreover, we also observed a signal to-noise ratio increase.

Suavization Results
Filtering seismic section using the proposed filter at crossline 905 of F3 block. Figure (b) shows a zoom area from the input data (a). Figures from (c) to (f) present the results obtained with 1, 3, 5 and 10 iterations, respectively. As we can see, although our method preserves structural features while denoising, for a large number of iterations these structures became smoother.
We implemented our solution using different numerical methods. We also compared 3 numerical libraries. We got the best result using GPU implementation of the Generalized minimal residual method. This result is up to 16x times faster than CPU for 10 iterations.

#CISBGF2017 #HPC #GPU #Oil&Gas #ComputationalGeophysics #AcceleratedGeophysics #SeismicInterpretation #SeismicFiltering #StructuralFeatures

@SBGf @Petrobras

Nenhum comentário

Deixe uma resposta

O seu endereço de e-mail não será publicado. Campos obrigatórios são marcados com *

Metta Innovations @ 15th CISBGf “Mapping Seismic Horizons using Clustering-based Methodologies”

At the 15th CISBGf (International Congress Of Brazilian Geophysics Society) that happened between July 31 and August 3 in Rio de Janeiro our principal researcher Aurélio Figueiredo gave a talk about “Mapping Seismic Horizons using Clustering-based Methodologies” during the Workshop on Computational Geophysics. A pioneer research in modern Machine Learning applied to Seismic Interpretation. This technique provides …

Deprecated: Directive 'track_errors' is deprecated in Unknown on line 0