To improve recommendation quality, the existing trust-based recommendation methods often directly use the binary trust relationship of social networks, and rarely consider the difference and potential influence of trust strength among users. To make up for the gap, this paper puts forward a hybrid top-N recommendation algorithm that combines mutual trust and influence. Firstly, a new trust measurement method was developed based on dynamic weight, considering the difference of trust strength between users. Secondly, a new mutual influence measurement model was designed based on trust relationship, in light of the social network topology. Finally, two hybrid recommendation algorithms, denoted as FSTA(Factored Similarity model with Trust Approach) and FSTI(Factored similarity models with... trust and influence), were presented to solve the data sparsity and binarity. The two algorithms integrate user similarity, item similarity, mutual trust and mutual influence. Our approach was compared with several other recommendation algorithms on three standard datasets: FilmTrust, Epinions and Ciao. The experimental results proved the high efficiency of our approach.