Abstract: A ultrasound signal deconvolution model in the framework of the sparse decomposition is proposed to improve the quality of medical ultrasound images. The smoothness of the signal and the sparsity of the dictionary representation are constrained by using two regularization terms, and the point spread function is estimated by using higher order statistics and MA model. The proposed model is solved by alternatively iterating split Bregman method. The gray scale ultrasound image is acquired by the dynamic filtering, envelope detecting, second sampling, dynamic compressing, and gray scale mapping. Experiments show that the proposed deconvolution method can achieve images with higher resolution, better contrast enhancement, and less speckle noise, compared with direct imaging methods.