Personalized recommendation is valuable in various web applications, such as e-commerce, music sharing, and news releasing, etc. Most existing recommendation methods require users to register and provide their private information before gaining access to any services, whereas a majority of users are reluctant to do so, which greatly limits the range of application of such recommendation methods. In the non-register environments, the only available information is the content or attributes of resources and the click-through chains of user sessions, so that many recommendation methods fail to work effectively due to the rating sparsity [Adomavicius and Tuzhilin, 2005] and illegibility of user identity, collaborative filtering [Goldberg et al. 1992] is an example of this case. In this paper we propose a joint recommendation method combining together two approaches, namely the domain category tree and the associate graph, to make full use of all available information. Further, an associate graph propagation method is designed to improve the traditional associate filtering method by integrating additional graphical considerations into them. Experiment results show that our method outperforms either the single category tree approach or the single associate graph approach, and it can provide acceptable recommendation services even in the non-register environment.