Micro-seismic source imaging shares the same theory as seismic imaging, they both require two main steps: 1.propagate the recorded data at the receivers from the surface back into the subsurface; 2.use the appropriate imaging condition to locate where micro-seismic events occour. Analogous to the data-domain matching filters used in least square migration, we propose to apply this non-stationary filter in micro-seismic imaging. We specifically invert for a filter that matches the original recorded data with data modeled from a estimated imaged source. By applying this filter on the recorded data prior to back propagation, we are effectively applying an approximate Hessian to the least-square fitting problem that will correct for illumination issues and provide better focusing. Using this approach we obtain good micro-seismic source imaging results for single source and multiple sources.