GNSS ionospheric tomography technique is capable to reconstruct the high-quality 3D ionospheric electron density (IED) images with a relatively low cost. We present a new parameterized approach for refining the voxel-based ionospheric tomography modeling. This approach is different to most other voxel-based techniques as they assume a homogeneous IED distribution in each voxel that is unreasonable for the tomography modeling. In this method, IED of any point within a voxel is determined via vertically exponential interpolation and horizontally inverse distance weighted interpolation from the IED values at the eight corners of that voxel. The parameterized tomography is tested with real data collected over the period of June 1-30, 2015 from 45 GPS stations in south China. The superiority of the new parameterized method is verified by comparison with the traditional nonparametric method. The new parameterized method outperforms the traditional method by 12%, 10%, 5% and 2% for vertical resolutions of 25 km, 50 km, 75 km and 100 km, respectively, in the self-consistency validation by GPS data. Such improvements are 20%, 24%, 22% and 16%, respectively, when assessed by the Swarm in situ IEDs. In terms of the vertical layer discretization, configurations using the resolution of 25 km generally performs better than the other three vertical resolutions. Overall, the parameterized method using a vertical resolution of 25 km achieves the best performance from the comprehensive comparisons with ionospheric data derived by GPS, ionosonde and Swarm satellites.