Block-Wise 3D Ultrasound Image Super-Resolution
Abstract
This paper addresses the problem of 3D ultrasound (US) single image super-resolution (SR), i.e., recover a high-resolution volume from its blurred, decimated, and noisy version. A new 3D US SR technique based on a linear forward model is studied by taking into account the axial variability of the point spread function (PSF) within a block-wise recovery process. The PSF is estimated using a recent algorithm along the axial direction with the assumption that it is isotropic in the other two spatial directions. By exploiting the linear image formation model, a cost function is constructed from a data fidelity term penalized by an $l_1$-norm regularization function imposing the sparsity of the solution. Numerical results show the efficiency of the proposed method when compared to the observed image.
Origin : Files produced by the author(s)