The study about the compressed sensing technology indicates that selecting sampling model according to the sparse base has great influence on the reconstruction results. In the Fourier space, there are huge differences between the reconstructions of the stellate sampling of polar coordinates and the random sampling. The spectral distribution of the Fourier space is analyzed by Fourier optical theory and the phenomena is explained theoretically. It is proposed that the matching degree between the sparse base and the sampling model may affect the reconstruction effect. In the wavelet space, a comparison experiment of reconstruction is taken with the uniform sampling and the random sampling, and the latter is better. Therefore, a method is provided for the practical application to reduce the sampling rate and improve the reconstruction effect by choosing suitable sampling models according to the sparse bases.