One application of electrocardiographic imaging (ECGI) is the localization of the origin of VT/PVC. This can be done using a single dipole assumed to be representative of the early electrical activity of the myocardium. Under clinical conditions, the measured signals are contaminated by various types of additive noise, which need to be eliminated without negative influence on the inverse solution. The experimental ECG signals were obtained from a Langendorf-perfused pig heart placed in a human-shaped torso tank, provided by the Electrophysiology and Heart Modelling Institute (IHU-LIRYC), France. Simultaneous measurements of 108 epicardial signals together with 128 signals on the tank were performed during 30s of RV pacing. The measured tank signal was processed in several ways. First, three methods of baseline drift removal (BDR1-3) were applied to the original signal. Next, BDR3 was applied to signal after high-frequency noise removal (HFR) processing. Finally, for each processing procedure, an averaged heart beat was calculated as the input for the inverse solution with a single dipole. The localization error (LE) with respect to the known true position of the pacing site was evaluated as well as the relative residual error (RRE) between the input body surface potential map (BSPM) and the BSPM reconstructed from the computed dipole. The inverse solution was computed from the first 20ms of the beat. It was assumed that the starting point was 95ms before the R-peak of the beat. When only BDR was applied to the signal the LE varied from 1.0 to 6.0cm and RRE from 23% to 43%. HFR before BDR3 improved the RRE (14% to 21%), while LE (0.2-1.5cm) was similar to BDR3. As the inverse method used needs to be applied to a very early time interval of activation it is more sensitive to proper selection of the BDR than to HFR method.