Internet of Multimedia Things (IoMT) brings convenient and intelligent services while also bringing huge challenges to multimedia data security and privacy. Access control is used to protect the confidentiality and integrity of restricted resources. Attribute-Based Access Control (ABAC) implements fine-grained control of resources in an open heterogeneous IoMT environment. However, due to numerous users and policies in ABAC, access control policy evaluation is inefficient, which affects the quality of multimedia application services in the Internet of Things (IoT). This paper proposed an efficient policy retrieval method to improve the performance of access control policy evaluation in multimedia networks. First, retrieve policies that satisfy the request at the attribute level by computing based on the binary identifier. Then, at the attribute value level, the depth index was introduced to reconstruct the policy decision tree, thereby improving policy retrieval efficiency. This study carried out simulation experiments in terms of the different number of policies and different policy complexity situation. The results showed that the proposed method was three to five times more efficient in access control policy evaluation and had stronger scalability.