Functional magnetic resonance imaging (fMRI) is usually based on acquisition of alternating series of images under rest and an activation task (stimulus). Brain activation maps can be generated from fMRI data sets by applying several mathematical methods. Two methods of image postprocessing have been compared: (i) simple difference of mean values between rest and stimulation, and (ii) Student's t-test. The comparison shows that the difference method is very sensitive to arbitrary signal fluctuations as seen mainly in large vessels (e.g., in the sagittal sinus), leading to insignificantly activated spots in brain activation maps. In contrary, Student's t-test maps show strongly reduced sensitivity for fluctuations and have the advantage of giving activation thresholds by setting significance levels. This allows the comparison of activation strength between patient collectives by using a grid overlay technique leading to an observer independent quantification of the stimulation effects. The method was able to reproduce previous findings of activation differences between healthy volunteers and schizophrenic patients. Moreover, a simple algorithm for the correction of slight head movements during the functional imaging task is presented. The algorithm is based on shifting the fMRI data set relative to a reference image by maximizing the linear correlation coefficients. This leads to a further reduction of insignificant brain activation and to an improvement in brain activation map quality.