Cardiac arrhythmias are often characterized by non-repetitive complex activation sequences. The properties of electrical activity in cardiac tissue, such as activation time (AT), can be accurately determined using optical imaging of electrical signals using voltage-sensitive dyes. The electrical AT of optical action potentials is known to accurately correlate with the time of the maximal derivative (dF/dt(max)) of the upstroke of ensemble averaged optical signals. The sensitivity of dF/dt(max) to noise is therefore problematic for mapping AT from nonrepetitive activity. Here we experimentally evaluated a novel method to determine activation times (AT) from non-averaged epicardial optical signals. The method depends upon the time delays of activation between adjacent pixels developing a global AT map, as opposed to measuring local AT from each pixel independently based on dF/dt(max). ATs from dF/dt(max) and global AT maps of non-ensemble averaged signals were correlated with averaged signals. Global ATs improved linear correlation coefficients and accuracy of AT patterns. Furthermore, global AT maps were significantly more robust at reproducing AT maps between consecutive stable beats. The proposed method of global AT mapping will potentially enable accurate mapping of non-repetitive propagation during arrhythmias.