Producing accurate thickness maps of corrosion damage is of great importance for assessing life in the petrochemical industry. Guided wave tomography provides a solution for this, by sending guided waves through the region of interest, then using tomographic imaging techniques to reconstruct the thickness map, importantly eliminating the need to take measurements at all points across the surface. However, to achieve accurate maps, the imaging algorithm must account for the way in which the guided waves interact with corrosion defects, and the complex scattering which occurs. Traditional approaches have exploited the dispersive nature of guided waves: a velocity map is produced from a dataset, then converted to thickness using the dispersion relationship. However, these relationships are derived for plates of constant thickness, which is not the case in the majority of defects, causing significant inaccuracies to exist in the images. This paper develops a more sophisticated inversion solution which accounts for the full-guided wave scattering, enabling more accurate images with resolution better than a wavelength, compared with two wavelengths previously. This is demonstrated with simulated and experimental data. The speed and stability of the algorithm in the presence of random noise and systematic errors is also demonstrated.