Use of penalty function or Lagrangian multipliers for constrained deconvolution

摘要

Two methods are presented to achieve constrained deconvolution. The first is based on the use of the classical penalty function in the optimization procedure. This approach allows the introduction of various kinds of constraints. The second is more particular, because it works only with some constraints, especially with a positivity constraint. In this case, the use of Lagrangian multipliers gives rise to a 'nice' formulation and good restoration can be performed.