Eddy current testing is used to inspect and analyze defects in metallic structures that can damage its future usefulness. However, this technique still needs to be improved. In this paper, a probe including one planar coil, one GMR sensor and a tiny permanent magnet is attached to a two dimension positioning system to evaluate a metallic plate with a linear defect. A magnetic field perturbation around the defect is measured and image processing techniques, in conjunction with an inverse problem algorithm, are used to obtain the eddy current density present in the metallic surface. From the eddy current map it is possible to determine the geometrical characteristics of the defect. To apply the inverse problem algorithm, regularization to suppress noise and to overcome instability effects are required. In this paper, Tikhonov and Total Variation Regularization are compared regarding their performance and obtained results.