A new separable 2-D finite impulse response (FIR) filter design technique with sparse coefficients is proposed. This technique includes two stages. In the first stage, the iterative reweighted 1 norm technique is utilized to design a sparse separable 2-D FIR filter based on some initial design. In the second stage, the greedy searching technique is utilized to make more sparse coefficients based on the first stage design. Simulation experiment validates the effectiveness of the proposed technique.