Online Social Networks (OSN) are widely adopted in our daily lives, and it is common for one individual to registerwith multiple sites for different services. Linking the rich contentsof different social network sites is valuable to researchers forunderstanding human behaviors from different perspectives. Forinstance, each OSN has its own group of users and thus, has itsown biases. Linked accounts can be a good calibration dataset toimprove data quality. This Entity Resolution (ER) problem is achallenge in the social network domain that many researchersattempt to tackle. In this paper we take advantage of spatialinformation posted in different social network sites and proposean efficient multiresolution mutual information approach to linkthe entities from those sites. The proposed method significantlyreduces the computing time by utilizing an iterative coarse-tofinemultiresolution approach, yet is robust in dealing with thesparsity of location data. The human location-wise behavior isalso discussed in deciding the resolution level. Public availableTwitter and Instagram data collected from their APIs are usedto illustrate the method, and the performance is evaluated bycomparing it with greedy mutual information approach.