Accurate determination of local blood flow in tissue using the Kety-Schmidt one-compartment model for freely diffusible tracers requires knowledge of the true arterial input function in tissue. Because measured input functions are usually delayed and dispersed with respect to true influx, a correction of the experimental input function is necessary. We describe a technique that uses a fast multilinear least-squares minimization procedure to determine simultaneously the dispersion, the blood flow and the partition coefficient as a function of delay. In this approach, a few multilinear fits are sufficient to determine the complete set of parameters necessary to describe the data. Because of the high speed of the procedure, dispersion effects may be taken into account on a pixel-by-pixel basis in calculating parametric images of blood flow and partition coefficient. The described procedure has been used at our institute for about 1 yr in more than 160 investigations and has proven well suited for routine use in a clinical environment.