Non-rigid registration of medical images based on local linear embedding and improved L-BFGS optimization

Non-rigid registration of medical images has become a challenging task in medical image processing and applications. In this paper, we propose a local linear embedding (LLE) and improved L-BFGS (limited-memory Broyden Fletcher Goldfarb Shanno) optimization based registration method. With abundant spatial information and good stability in noisy environment, the ordinal features are computed on different orientations to represent spatial information in medical images. For high dimensional ordinal features, the LLE algorithm is used for dimensionality reduction and the inverse mapping of LLE is used to fuse complementary information together. Then a hybrid entropy based similarity measure which integrates image intensity with ordinal feature is chosen as the registration function. Finally an improved L-BFGS algorithm is used to search for the optimal registration parameters. We evaluate the effectiveness of the proposed approach by applying it to the simulated brain image data. Experimental results show that the proposed registration algorithm is less sensitive to noise in images. Compared with some traditional methods, the proposed algorithm is of higher precision and better robustness.