Structure-Preserving Smoothing for 3D Seismic Data Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing seismic data processing methods, such as Gaussian and mean filters, often smear structural edges and fail to preserve planar features in 3D datasets, while structure-oriented filtering is computationally costly and inaccurate in noisy regions.
Innovation Solution
A data adaptive structure-preserving smoothing method that tests smoothing in multiple orientations and selects the best result based on minimum deviation or other criteria, applicable to both structured and non-structured areas without prior computation of structural orientation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If Gaussian and mean filters are used for smoothing, then random noise is reduced, but structural edges and planar features are smeared and lost
Solution Approach 1:
The patent applies different smoothing operations to different local regions based on structural characteristics. By evaluating variance in multiple orientations and selecting the orientation with minimum variance, the method adapts the smoothing strength locally to preserve edges while reducing noise in homogeneous regions.
Solution Approach 2:
The patent dynamically selects smoothing parameters and orientations based on local data characteristics rather than using fixed parameters. The adaptive selection of orientation and smoothing window size allows the filter to respond to local structural variations, maintaining edges while reducing noise.
2Manufacturing precision
If structure-oriented filtering is used to preserve structural orientation, then planar features are maintained, but computational cost increases and accuracy decreases in noisy regions
Solution Approach 1:
The patent computes structural orientation only partially by evaluating a limited set of predefined orientations (e.g., 0°, 45°, 90°, 135°) rather than computing the exact orientation. This partial computation significantly reduces computational cost while still capturing the dominant structural directions for filtering.
Solution Approach 2:
The patent uses simple statistical measures (variance, standard deviation) as proxies for complex structural orientation analysis. These computationally inexpensive metrics provide sufficient information for adaptive filtering without requiring expensive iterative optimization or precise orientation computation.
3Manufacturing precision
If structure-oriented filtering is applied, then structural preservation is improved, but the method becomes inaccurate in noisy regions and fails in non-structured areas
Solution Approach 1:
The patent performs preliminary evaluation of variance in multiple orientations before applying the final smoothing operation. This preliminary assessment allows the method to identify regions with strong structural signals versus noisy or non-structured regions, and adapt the filtering accordingly to maintain reliability across different data types.
Solution Approach 2:
The patent creates a universal smoothing method that works for both structured and non-structured areas by using orientation-independent variance evaluation. The same algorithm adapts to different data characteristics (coherence, curvature, amplitude) without requiring separate processing paths, achieving multi-functionality and broad applicability.
Data Source
AI summary
A system and method perform structure-preserving smoothing (SPS) using a data adaptive method for smoothing 3D post-stacked seismic attributes which reduces random noise while preserving the structure without prior computation of its orientation. The data is smoothed within a neighborhood sub-window along a set of predefined orientations, and the best smoothing result is then selected for output. The orientation corresponding to the best result often approximates the true structure orientation embedded in the data, so that the embedded structure is thus preserved. The SPS method can also be combined with median, alpha-trim, symmetric near neighbor, or edge-preserving filters. The SPS method is an effective way to reduce random noise and eliminate noise footprints, and to enhance coherence and curvature attributes.