The shift of the Internet usage from a host to host interconnection to a content retrieval becomes an indisputable fact. Information Centric Networking (ICN) paradigm is proposed as a Future Internet architecture adapted to this change. By identifying the content and not the host, this new design enables many interesting features such as in-network data storage and caching. ICN network components will be able to recognize transiting contents, store them and serve efficiently new dissemination demands without contacting the original data providers. This feature affects surely the ICN performance. In this paper, we aim to deal with the ICN caching. We introduce POPCAM, a new strategy based on three aspects: an off-line popularity estimation, a placement decision with variable proximity control and a replacement technique combining LUV(Least Unified Value) with the measured popularity distribution. Our strategy adapts a caching decision to the resolution topology and the replacement technique to the content popularity estimation. Evaluation results reveal that POPCAM is able to achieve a high hit rate.