In this paper, we propose a multi-antenna frequency-shift (FRESH) filter structure which combines the time-dependent cyclic Wiener filter with spatial filtering. The new solution can be applied as a blind technique, which is desired in many applications where training signals are not available. When compared to a traditional space-time adaptive processing (STAP) architecture, our new method shows the capability of providing the same performance, but with a smaller number of antennas. This reduction in the number of array elements is critical for those applications where there are serious limitations on the size of the array, e.g., portable devices. We have also reduced the computational complexity of the multi-antenna FRESH filter, by using the multistage Wiener filter rank reduction technique. Simulation results are provided to demonstrate the usefulness of the proposed interference canceller.