The channel mismatch problem was investigated in multi-aperture Synthetic Aperture Radar (SAR). A channel balancing technique based on the two-stage filters in the different raw data domains was presented. The amplitude and phase filter in the range Doppler domain corresponds to the antenna beam pattern difference, and the filter in the range frequency azimuth time domain corresponds to the frequency response difference. In the first stage filtering process, we constructed an adaptive filter in the region with higher coherence, and a non-adaptive one in the noise dominated region to remove the deterministic difference between channels. In the second stage filtering process, the noise was removed and only the target signatures were left. The feasibility of the channel balancing algorithm carried out in the raw data domain is demonstrated with the measured airborne data.