Building Blocky Models With Full Waveform Inversion

Deriving blocky models from Full Waveform Inversion (FWI) might be required in some geological settings. These models might be also needed for imaging with wave-equation based techniques such as Reverse Time Migration (RTM). Enforcing blockiness in the model space amounts to enforcing sparseness in the derivative of the model. Sparseness can be obtained by using long tail distributions such as the l1 norm or Cauchy function. By using different directions for the derivatives, the model can be made blocky in more than one direction: horizontally, vertically, or both. Whereas the Cauchy function produces better blocky models, the l1 norm has the advantage of being convex and easy to parameterize.