Gimbaled imaging systems require very high performance inertial stabilization loops to achieve clear image acquisition, precise pointing, and tracking performance. Therefore, higher bandwidths become essential to meet recent increased performance demands. However, such systems often posses flexible dynamics around target bandwidth and time delay of gyroscope sensors which put certain limit to achievable bandwidths. For inertial stabilization loops, widely used design techniques have difficulty in achieving large bandwidth and satisfying required robustness simultaneously. Clearly, high performance control design hinges on accurate control-relevant model set. For that reason, combined system identification and robust control method is preferred. In the system identification step, accurate nominal model is obtained, which is suitable for subsequent robust control synthesis. Model validation based uncertainty modeling procedure constructs the robust-control-relevant uncertain model set, which facilitates the high performance controller design. Later, with skewed-mu synthesis, controller is designed which satisfies large bandwidth and robustness requirements. Finally, the experimental results show that significant performance improvement is achieved compared to common manual loop shaping methods. In addition, increased performance demands for new imaging systems are fulfilled.