New techniques for improving both the computational andimaging performance of the three dimensional (3D) electromagnetic inverseproblem are presented. A non-linear conjugate gradient algorithm is theframework of the inversion scheme. Full wave equation modelling forcontrolled sources is utilized for data simulation along with anefficient gradient computation approach for the model update. Improvingthe modelling efficiency of the 3D finite difference method involves theseparation of the potentially large modelling mesh, defining the set ofmodel parameters, from the computational finite difference meshes usedfor field simulation. Grid spacings and thus overall grid sizes can bereduced and optimized according to source frequencies and source-receiveroffsets of a given input data set. Further computational efficiency isobtained by combining different levels of parallelization. While theparallel scheme allows for an arbitrarily large number of parallel tasks,the relative amount of message passing is kept constant. Imageenhancement is achieved by model parameter transformation functions,which enforce bounded conductivity parameters and thus prevent parameterovershoots. Further, a remedy for treating distorted data within theinversion process is presented. Data distortions simulated here includepositioning errors and a highly conductive overburden, hiding the desiredtarget signal. The methods are demonstrated using both synthetic andfield data.