The analysis of instantaneous and average rolling leukocyte velocity is crucial to the study of inflammatory disease. In order to record features associated with leukocyte rolling, the leukocyte position must be tracked, typically by manual observation. Automated tracking of leukocytes is possible for in vitro studies, but not for recordings resulting from intravital experiments. Therefore, we have designed and implemented an image processing system for automated tracking of rolling leukocytes in vivo. The novel image processing techniques used in the tracking system successfully address the four major problems associated with tracking cells in vivo: background movement, severe image noise and clutter, cell deformation and contrast change, and occlusion of the target cell by other structures. We have tested the system in two experimental protocols in which leukocyte rolling is observed in venules of the mouse cremaster muscle with and without TNF-alpha treatment. The automated tracking system was validated by comparing automatically generated displacement and velocity data with data from the same recordings collected manually. The root mean squared error between the computed displacements and the manually measured displacements was less than 12% of the average displacement in TNF-alpha-treated venules. The average velocity error was also less than 12%. For untreated venules, the computed and measured displacements and velocities had an RMSE of less than 8%. The automated tracking system allows one, for the first time, to reliably track rolling leukocytes in vivo, thus eliminating possible investigator bias and increasing throughput.