We propose a new inversion approach to estimate Q factor values, based on back propagating the seismic data to its source location. If the correct velocity and Q-factor models are used, the back-propagated frequency spectrum will be consistent in all receivers recording the same source. The Therefore, we formulate a corresponding objective function inversion problem that is solved in a global optimization scheme. This workflow doesn't require any knowledge of the source wavelet or an initial Q-factor model. In addition, it is less sensitive to noise as it takes multiple channels into account in a large frequency band. While this approach is conceptually true for any back-propagation method, we construct and show its application using the linearized ray-based approximation. Through synthetic and field data examples, we demonstrate the application and the usefulness of the suggested workflow.