A Regularization Algorithm Optimized for Time-lapse Processing

In time-lapse processing, independent regularization of each vintage is the typical approach. This disregards any geometrical limitations imposed by different surveys. Here we recast the regularization process as a minimization problem with model-space constraints. These constraints couple geometrical relations between surveys to improve repeatability. We also demonstrate how to solve the minimization problem using a practical and pragmatic approach. Results from a North Sea dataset show overall reduction of 4D noise, especially around less repeatable parts of the surveys including the undershoot zone.