A method to automatically segment the hippocampus and the amygdala using multi-channel large deformation diffeomorphic metric mapping (LDDMM) is proposed. T1 data from normal young subjects were used to measure the segmentation accuracy. To examine the impact of morphological abnormalities on the accuracy, the method was also tested using subjects with Alzheimer's disease (AD). The segmentation accuracy was compared with two state-of-the-art algorithms - FSL and Freesurfer. To improve the segmentation accuracy, LDDMM cascading was adopted. For normal subjects, the mean kappa overlap ratios of the automated segmentations with the manual segmentations were 0.76 and 0.84 for the hippocampus and the amygdala, respectively. For AD patients, the respective mean kappa overlap ratios were 0.76 and 0.8.