Edge-preserving smoothing for segmentation of seismic images

Image segmentations

Finally, we should examine the effect of MH filtering on automatic segmentation results. Figure 10(a) is the result of segmenting the original, unfiltered image in Figure 6. ``Leakages'' are apparent from the salt body into the surrounding sediments, especially near the intersecting location indicators. When the MH-filtered image is segmented, however, these problems are greatly ameliorated (Figure 10(b)). A second field data example is the attempt to segment the small salt inclusion indicated in Figure 11. While segmenting the original image leads to the poor result in Figure 12(a), segmenting an MH-filtered image provides a much improved result (Figure 12(b)). In these examples, hybrid-MH filtering has allowed for more accurate segmentation results; furthermore, the computational efficiency (and simple parallelization) of the algorithm make it especially attractive since smoothing the image in these examples required only a fraction of the time needed for the already-efficient PRC segmentation algorithm.

seg-orig,seg-mhn
Figure 10. Automatic image segmentation results when using (a) the original, unfiltered image in Figure 6, and the MH-smoothed image in Figure 9(d).

oct-full
Figure 11. Another image from the Gulf of Mexico featuring an inclusion within the salt body. Further examples will be shown from the indicated area.

oct-orig-seg,oct-smth-seg
Figure 12. Automatic segmentation results using (a) the original, unfiltered image and (b) an MH-smoothed image.

Edge-preserving smoothing for segmentation of seismic images