Accurate registration of PET and CT images is an important component in oncology, so we aim to develop an automated registration algorithm for PET and CT images acquired by different system. These two modalities offer affluent complementary information: CT provides specificity to anatomic findings, and PET provides precise localization of metabolic activity. In this paper, we proposed an improved registration method that can accurately align PET and CT images. This registration algorithm includes two stages. The first stage is to deform PET image based on B-Spline Free Form Deformations (FFD). It is consists of three independent steps. After the feature points of PET and CT images have been extracted in the preprocessing step. As a next step, the PET image is deformed by B-Spline Free Form Deformations (FFD) with feature points of CT images. The second stage is to register PET and CT images based on Mutual Information (MI) of feature points combined with Particle Swarm Optimization (PSO) algorithm and Powell algorithm that are used to search the optimal registration parameters.