Assessment of iron overload in liver by T2* magnetic resonance imaging (MRI) is a widely used clinical procedure. In the common clinical practice, measurement is performed locally by manually drawing a small region of interest in liver. This procedure may be affected by a noticeable intra- and inter-observer variability. In this study, a new approach is proposed that performs a global semiautomatic measurement of T2* involving the whole liver extension. Parenchyma is automatically segmented by an adaptive fuzzy-clustering algorithm. The liver T2* global value is evaluated using a pixel-wise approach by introducing an appropriate signal decay model. The proposed method was tested on a synthetic software model and on MR images acquired from 30 thalassemia major patients. The methods was demonstrated to increase the measure precision in T2* assessment and to significantly reduce the intra- and inter-observer variability.