Abstract: Due to conventional sparse representation bases are not targeted, so the application in synthetic transmit aperture medical ultrasound imaging is not efficient, and it's hard to guarantee the reconstruction effect with a low compression ratio, A high efficient sparse dictionary is designed to solve the problem. Based on the properties of ultrasound echo signals, the dictionary uses transmitted pulse as the basis function, sparse representations of echo signals on the proposed sparse dictionary is very sparsity and theoretically the sparsity of sparse coefficients can be equal to the number of reflection signals. Simulations of the simple point targets and the complex targets through the simulation tool Field Ⅱ demonstrate that the mean absolute error using the sparse dictionary is obvious smaller with the same reconstruction algorithm and compression ratio compared to conventional sparse representation, it is a fraction of DWT and it is a few one-tenth and even a few percent of DCT and DFT. The sparse dictionary can reach the same recovery effect by a lower compression ratio. Finally, an experiment is conducted on the basis of the actual data. The experimental result is also in line with the simulation. The compressed sensing method based on the proposed sparse dictionary can furtherly reduce the complexity of system and the amount of acquired echo data needed to reconstruct the ultrasound imaging