Data Filtering by 3D Convolution

@inproceedings{TertoisRM2004,
 abstract = { Two-dimensional convolution is a common technique for image processing. Different types of convolution kernels are used depending on the desired filtering effect. The same principle can be applied in a G CAD plug-in to data volumes. Applying a three-dimensional convolution kernel makes it possible to take into account the correlation of the data in three dimensions. A smoothing convolution kernel can be used as a local estimator. For data sets in which some values are undefined, an approximation of the value can be computed on a local neighbourhood by convolution with an averaging kernel. Applying a three-dimensional convolution kernel to data volumes can be done simply by iterating on data and kernel values. For seismic data volumes, filtering by a chosen type of kernel can facilitate interpretation. Sharpening a coherence data set can enhance faults and make them easier to pick. Smoothing an amplitude data set can improve auto tracking performance. },
 author = { Tertois, Anne-Laure AND Frank, Tobias },
 booktitle = { 24th gOcad Meeting },
 month = { "june" },
 publisher = { ASGA },
 title = { Data Filtering by 3D Convolution },
 year = { 2004 }
}