Differential evolution (DE) is one of the highest performance, easy to implement, and low complexity population-based optimization algorithms. Population initialization plays an important role in finding better candidate solution and faster convergence of the population to a global optimum. It has been shown in the literature that large population sizes for large-scale problems necessarily does not show a statistically significant performance improvement over medium size population. In this paper, we emphasise on importance of population initialization and discuss effects of using centroid-based population initialization in DE, with focus on micro-DE (i.e. DE with small population size). Experimental results for high and low dimensional problems with small and standard population sizes on CEC Black-Box Optimization Benchmark problems 2015 (CEC-BBOB 2015) show centroid initialization can increase performance of DE algorithm, compared to the conventional initialization method.