Iterative Dip-Steering Median Filter for Seismic Noise
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing seismic data processing techniques, such as median filtering and frequency-wavenumber filtering, struggle to effectively attenuate random noise in seismic data with conflicting dips, often distorting signals or failing to preserve edges.
Innovation Solution
The method involves transforming seismic data into the frequency-wavenumber domain, then the Fourier-radial domain, identifying dominant dips, and applying a median filter along these dips in a descending sequence to attenuate noise, allowing for effective noise reduction while preserving signal integrity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If median filtering is applied to seismic data, then random noise is attenuated, but it cannot handle data with conflicting dips and distorts signal levels
Solution Approach 1:
The patent applies iterative dip-steering median filtering that dynamically adapts to conflicting dips by repeatedly updating dip estimates and reapplying the filter. The method iteratively refines dip measurements and adjusts filtering directions to handle multiple dip directions in the seismic data, making the filter adaptable to complex geological structures with conflicting dips.
Solution Approach 2:
The patent changes the filtering parameters by computing dip-specific median filters with different dip angles and weights. By varying the dip parameter across iterations and applying weighted combinations of dip-specific filters, the method effectively handles conflicting dips while preserving signal integrity.
2Object-affected harmful factors
If frequency-wavenumber filtering is used, then random noise is suppressed, but signal levels are significantly distorted when strong noise exists
Solution Approach 1:
The patent computes multiple dip-specific median filters with different dip parameters and combines them using weighted sums. By changing the dip parameter and applying optimal weights, the method achieves effective noise suppression while preserving accurate signal levels, avoiding the distortion problems of conventional f-x filtering.
Solution Approach 2:
The patent creates a composite filtering approach by combining multiple dip-specific median filters into a weighted sum. This composite filter leverages the strengths of individual dip-specific filters while compensating for their limitations, achieving both noise suppression and signal level preservation.
3Manufacturing precision
If conventional median filtering is applied, then edge preservation is achieved, but it assumes flattened events and cannot handle conflicting dips
Solution Approach 1:
The patent extends conventional median filtering by making it dynamic and adaptive to dip variations. The iterative dip-steering approach continuously updates dip estimates and adjusts filtering directions, enabling edge preservation in non-flattened events with conflicting dips without requiring pre-flattening.
Solution Approach 2:
The patent segments the filtering process into multiple dip-specific components, each handling a specific dip direction. By dividing the complex filtering problem into manageable dip-specific segments and combining them, the method preserves edges while handling conflicting dips in non-flattened events.
Data Source
AI summary
An iterative dip-steering median filter is provided for random noise attenuation in seismic data where conflicting dips are indicated in the data. A number of dominant dips inside a processing window or sample of the data are identified by a Fourier-radial transform in the frequency-wavenumber domain. A median filter is then applied along the dominant dip to remove noise, and the remaining signal after filtering is retained for further median filter iterations. Iterations are repeated to apply the median filter along the most dominant dip in the remaining data. The processing continues in subsequent iterations until all selected dips have been processed. The remaining signal of each iteration is then summed for final output.