f 2D Signal and noise decomposition for sparsely and irregularly-sampled data

A general method for separating signal from noise is to apply a transform to a subset of the data (often a CMP gather) which maps signal and noise on separate parts of the transform domain. The transform should differentiate between signal and noise. Several commonly used transforms are based on the difference in moveout (as a function of offset) between signal and noise. Examples are k -f filtering, linear- and parabolic Radon transform (RT). Many processing algorithms based on these transforms require regularly sampled data along the offset direction. However, in 3D seismics, the data is often irregularly and, even worse, sparsely sampled.