Abstract: Costas-coded Frequency Hopping (FH) signal is very sensitive to velocity changes. As a result, in multi- targets scenario, sidelobe pedestal arisen induced by inaccuracy compensation will make possible masking if one target is much stronger than the others. Based on this property of Costas-coded FH signal a high resolution rang- velocity two-dimensional imaging algorithm is studied by using sparse component analysis. Echoes can be regarded to be sparse in rang-velocity dimensions, and a sparse dictionary is constructed based on the sparsity feature, where the waveform entropy is used as the sparse measuring function. By improving the generalized regularized FOCUSS algorithm and making use of Kalman filter, the best atom can be found iteratively and high resolution in rang and velocity is realized. The algorithm is validated by simulations, and the noise tolerance as well as the time effectiveness are analyzed.