Abstract: Compressed sensing provides a new technique for solving an underdetermined system of linear equations according to sparse nature of its solutions. In this paper, the SAR side-lobe reduction technique based on compressed sensing is proposed. Under the high signal to noise ratio condition, the side-lobe can be suppressed at no cost of resolution of the image, and super-resolution can be also achieved when the scatters are sufficiently sparse; At low SNR is not only a better suppression of the side-lobe, but also inhibit the effect of noise. One-dimensional, two-dimensional and actual data processing verified the validity of this algorithm.