Seismic data can experience non-stationary frequency variations caused by attenuation. This problem is encountered when matching multiple data sets, such as in multicomponent image registration, because signals with di˙ering frequency content are hard to correlate. In this paper, we propose a method to balance frequency content between data sets while taking into account non-stationary frequency variations. This method involves finding and applying a non-stationary smoothing operator to minimize the local frequency di˙erence between data sets. Numerical examples demonstrate that the proposed method improves multicomponent image registration and matching images of differing resolution.