Edge-preserving smoothing for segmentation of seismic images

Conclusions

Automated seismic image segmentation schemes can benefit from preprocessing such as smoothing prior to segmentation. An edge-preserving smoothing method based on directional maximum homogeneity can de-noise and clean up an image, while preserving important edges such as salt boundaries. Applying the MH filtering algorithm to 3D field seismic data improves segmentation results by reducing ``leakages'' through boundaries. In addition, a hybrid approach combining the best aspects of traditional smoothing and MH filtering can improve noise removal while still preserving edges well.