Photo-acoustic Tomography (PAT) is a hybrid technique, which takes advantage of high optical contrast provided by optical imaging and rich spatial resolution attributed to ultrasound. The Time reversal (TR) method is the most popular strategy for solving the inverse problem in PAT, thanks to its ability to deal with inhomogeneous properties of the medium or irregular surface shapes. The performance of the TR is, however, dependent on the well-posedness of the problem, and is deleteriously affected by problems such as data incompleteness or noise in data. For such cases, iterative optimization techniques are needed. However, in practice the application of iterative methods to PAT is very limited since the arising forward and adjoint operators are very costly. Recently, a TV-regularized version of FISTA has shown promising results in PAT, thanks to its better rate of convergence. In this seminar, we show that its performance on the limited-view PAT is greatly improved by a multi-grid scheme.