The image registration is a process of establishing spatial correspondences between two images. It is widely used in the computer vision, the remote sensing data analysis and the image processing. Especially in the image-guided radiation therapy, the image registration plays an important role. Recently, the scale-invariant feature transform (SIFT) has been used in the medical image registration, and obtained promising results. However, SIFT is apt to detect blob features which cannot reflect properly motions of lungs. In this paper, a hybrid feature detection method, which can detect lung tissue features effectively based on Harris and SIFT algorithms, was proposed. In addition, a novel method which can remove mismatched landmarks was also proposed. A series of thoracic CT images were tested by using the proposed algorithm. The quantitative and qualitative evaluations show that our method is much better than the conventional SIFT method especially in the case of large deformations of lungs during the respiration.