Deblending using normal moveout and median filtering in common-midpoint gathers

Median filtering after NMO

By utilizing the effectiveness of median filtering in removing spike-like noise and inserting median filtering into a common seismic processing workflow, we propose the following new processing workflow:

1. Transform the blended data from CSG to CMG.
2. Apply velocity scan and pick the NMO velocity.
3. Apply normal moveout.
4. Apply median filtering along the offset direction in CMG.
5. Apply inverse normal moveout.
6. Prestack migration in CMG or transform back to CSG for other processing tasks.

The key point in the above workflow is flattening the seismic events in order to apply median filtering, which results from obtaining a convincing velocity scan. Because of the intense blending noise, the velocity scan could not be obtained in one step. To solve the problem, we may need to implement steps 2-4 recursively in order to get a better velocity scan. Two velocity scanning iterations are usually adequate to get an acceptable velocity. The proposed processing flow can recursively polish both deblended result and velocity estimation. On one hand, the better velocity estimation can help to make the NMO-corrected events flatter, which improve the performance of median filtering to remove blending noise and to preserve useful energy. On the other hand, the better deblended result can also help to improve the velocity estimation.

Deblending using normal moveout and median filtering in common-midpoint gathers