Today, we compare popular post-stack noise reduction filters in OpendTect. The data set is Penobscot, a free data set that is downloadable from TerraNubis, our marketplace for free and commercial data sets. A special feature of Penobscot is that it allows users to run all commercial plugins without the need for a license key. This means that anyone can replicate the tests shown in this post.
We applied 3 noise reduction filters and compare these with the input data (PSTM processed) on one inline and one crossline. The 3 filters are:
- Dip-Steered Median Filter with a stepout of 2
- AJAX, a trained Machine Learning model (3D Unet) from AkerBP (Lundin-Geolab) available in OpendTect’s library of pre-trained models that has learned to reduce noise, enhance lateral continuity, and optimize frequency content.
- DeSmile, another trained Machine Learning model (3D Unet) from AkerBP (Lundin-Geolab) available in OpendTect’s library of pre-trained models. DeSmile has learned to reduced migration smiles artefacts.
Share your thoughts in the comments below and let us know which filter you prefer!