Knowing the dielectric constant of a geologic medium is an important task in GPR studies. Accurate estimation allow realistic results in migration process, whilst inaccurate values lead to incorrect time-to-depth conversion of GPR sections. Dielectric constant is also needed for target imaging through Microwave Tomography, which is an inverse scattering problem. In this paper an approach for dielectric permittivity evaluation through the tomographic imaging process in order to retrieve the best image possible for a given target is proposed. Input parameters required for the tomographic imaging process were evaluated through Ant Colony Optimization, an efficient global search algorithm. An experiment was carried out using a 2600 MHz antenna for surveys over metallic and Styrofoam targets of different cross sections embedded in a sand box, in five different scenarios. Accurate results were achieved for estimating dielectric permittivity, with slight variations depending on the scenario complexity. True target positions were well retrieved in most cases; however, depth and shape estimation presented higher errors related to known limitations of the tomographic imaging. In addition, automatic estimation of the regularization factor avoids eventual errors related to subjective analysis. Low computational times required to retrieve the models from real data makes this approach suitable for application in realistic GPR surveys.