In signal processing, the autocorrelation time can mean the time it takes for a signal to 'lose it's memory'. For example, plotting the position of a diffusing particle over time and autocorrelating the signal gives a short-time scale where the position is non-random as compared to previous positions, and a longer time scale where the current position is not correlated with earlier positions. Autocorrelation is used a lot in light scattering experiments as well- it gives a scale for the dynamics of the system. Also in optical processing techniques such pattern recognition.
Covariance is a more general way of discussing the coherence properties of *two* signals, but probably reduces to the autocorrelation when the signals are from the same source.