This paper addresses the problem of blind deconvolution of medical ultrasound (US) images. Specifically, a parametric model for the point spread function (PSF) established experimentally is used, i.e., the US PSF can be modeled by a Gaussian function modulated by a sinusoidal function. Given this parametric model, the estimation of the PSF in a blind deconvolution problem can be reduced to the estimation of its parameters. Moreover, due to the ill-posedness of blind deconvolution problem, an ℓp-norm (0 <; p ≤ 2) regularization term (including the widely considered ℓ1-norm, ℓ2-norm regularization terms) for the ultrasound tissue reflectivity function (TRF) is employed, based on the assumption of generalized Gaussian distributed US images. An alternating optimization approach is proposed for the estimations of the US PSF and TRF. The behavior of the proposed algorithm is illustrated using simulated and in vivo US data.