Subset-based local DIC and finite element-based (FE-based) global DIC are the two primary image matching methods that have been extensively investigated and regularly used in experimental mechanics community. Due to its straightforward implementation and high efficiency, subset-based local DIC has been used in almost all commercial DIC packages. However, it is assumed by some researchers that element-based global DIC offers better accuracy because of the enforced continuity between element nodes. Thus there is a pressing need to comprehensively examine the performance of these two DIC approaches. In this work, theoretical analyses of the standard deviation errors of classic subset-based DIC and two FE-based DIC techniques are first performed. Then, by measuring displacements of the same calculation points using the same calculation algorithms and identical calculation parameters, the performances of subset-based local DIC and two FE-based global DIC approaches are compared experimentally in terms of measurement error and computation efficiency using numerical tests and real experiments. A detailed examination of both the theoretical and experimental results reveals that, when subset (element) size is not very small, standard subset-based local DIC approach not only provides better results in measured displacements, but also demonstrates much higher computation efficiency. However, several special merits of FE-based global DIC approaches are indicated.