Grid resilience indicates the ability of a system to withstand extreme disasters and return to normal operation. To improve the level of grid resilience assessment, a comprehensive evaluation method that takes into account the operation processes of the power grid is proposed. By decomposing the resilience curves into outage and restoration processes, this method reveals the process of extracting resilience indicators from standard power system data. Firstly, resilience curves are generated from historical grid data, which are decomposed into outage processes and recovery processes. These processes usually overlap in time. Next, the resilience events in the data are used to characterize the statistics of these processes. The resilience assessment indicators, such as restoration duration, customer outage hours, and outage and restoration rates, are derived from these statistical data. Then, the means of these indicators are expressed as functions of outage number. Subsequently, a variability model for the restoration duration is developed that can predict the maximum restoration duration with 95% confidence level. Finally, the performance of the proposed method is analyzed numerically. The results validate its effectiveness.