With the proliferation of the diversified applications on heterogeneous media, online contents and offline services have influenced the daily life of web users in a more comprehensive way. In this paper, the concepts and methodologies for heterogeneous media analysis, which aim to obtain physical and social attributes from huge amount of data from heterogeneous sources were introduced, and the intrinsic mechanism for modeling the semantic divergence, complicated correlations and the information propagation was revealed. Firstly, we discussed the heterogeneous media with properties of cross-platform, cross-modality and diversity, and put forward the challenges and opportunities in studying heterogeneous media analytics, and discussed the impact of studying the heterogeneous media. Secondly, we provided literature review from three perspectives, i.e., semantic analysis and understanding, heterogeneous correlation modeling and community modeling. Lastly, we introduced our research works on the semantic modeling, correlation modeling and social behavior modeling on heterogeneous media resources.