Related Experiment Videos
Random noise attenuation using dual-module attention with multi-scale kernels.
Jayant Jharkhande1, Rajesh R Nair2
1Petroleum Geomechanics Laboratory, DOE, IIT Madras, Chennai, India.
Scientific Reports
|July 3, 2026
Summary
This study introduces a new Multi-Kernel Channel-Spatial Attention (MKCSA) module to reduce random noise in seismic data. The MKCSA-Net significantly improves seismic data quality for better geophysical exploration.
Area of Science:
- Geophysics
- Artificial Intelligence
- Signal Processing
Background:
- Improving seismic data quality is crucial for exploration geophysics.
- Random and coherent noise degrade seismic records, hindering interpretation.
- Conventional noise reduction methods can remove valuable geological signals.
Purpose of the Study:
- To develop a novel attention module for effective random seismic noise attenuation.
- To enhance the feature extraction capabilities for complex geological structures.
- To improve the signal-to-noise ratio (SNR) in seismic data.
Main Methods:
- Introduction of the Multi-Kernel Channel-Spatial Attention (MKCSA) module.
- Integration of multi-kernel feature extraction with channel and spatial attention mechanisms.
- Enrichment of the network's feature space for complex geological feature extraction.
Main Results:
- The MKCSA-Net effectively attenuates random seismic noise.
- The proposed attention mechanism enhances the extraction of local and global geological features.
- Significant improvements in Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), and Frequency Distance (FD) were observed compared to benchmark models.
Conclusions:
- The MKCSA-Net demonstrates superior performance in seismic data denoising.
- The novel attention module effectively suppresses noise while preserving geological signals.
- This approach offers a promising solution for enhancing seismic data quality in exploration geophysics.