Our knowledge of preferential flow in heterogeneous unsaturated porous media such as soils is still limited. In this work, we introduced a novel method based on time resolved three-dimensional images fast acquisitions using a four-slice helical computed tomography (CT). This method provides new quantitative information on preferential flow processes occurring in an unsaturated undisturbed soil. Twenty-four three-dimensional images were acquired every 3 min in transient flow regime and every 5 min in stationary flow regime during a specific simulated rainfall event hold inside the scanner gantry. The macroporosity thresholding was realized according to a linear combination of attenuation coefficients of the main soil phases (“mobile” air or water in macropores and saturated soil matrix) weighted by their volume fractions in voxels. Mobile water contained in the macropores and in the surrounding thinner pores (diffuse macroporosity) was identified using a specific thresholding method based on subtracted images. In spite of the experimental limitations—noise, artifacts, limited spatial and temporal resolutions that are discussed in the paper—the method allows identifying about 65% of the infiltration water present in the core during the rainfall event. Although identifying water films on images is not possible when their volume is lower than the detection threshold, quantitative time-lapse slice-averaged information (fraction of active macropores and mean water filling of active macropores) can be obtained. We showed that even during a quite intense rainfall event, macropore flow was localized only into 30% of the macropore network of the soil core; that is, macropores remained mostly unsaturated along the experiment.