Microtubules are tubular biopolymers of the cytoplasm. They play numerous critical roles in a cell such as providing mechanical support and structural tracks for the anchoring and transport of chromosomes, organelles, and vesicles. They also form the microtubule assembly, which is critical for the coordination of mitosis and cell migration. The most dynamic part of the assembly are the microtubule outer, plus, tips located close to the cortex of the cell. Abnormal function of the assembly has been implicated in cell pathology such as neurodegenerative diseases and cancer. To date the study of the dynamics of the microtubule assembly is often performed qualitatively by visual inspection or quantitatively by manual annotation of the locations of the tips over time in an image sequence, which is very tedious. In this work we have developed a method to automatically track microtubule tips so as to enable a more extensive and higher throughput quantitative study of the microtubule assembly. Our approach first uses the entire image sequence to estimate the region in which a tip oscillates. In that region a tip feature is computed for all time and subsequently used to form the tip trajectory. Last, we evaluate our method with phantom as well as real data. The real data show fluorescently tagged living cells imaged with epifluorescent microscopy or confocal microscopy.