For labeling the T2-weighted MR images using human brain atlas, it is prerequisite to the foundation of the Talairach space for T2W MR images, and the basic condition to found Talairach space is the location of Talairach cortical landmarks from T2W MR images. A method to locate the Talairach cortical landmarks from T2W MR images is proposed, it consists of three aspects: Firstly, determine the planes including the six cortical landmarks ; segment the planes based on fuzzy C-means clustering algorithm, gray level projection, watershed algorithm, region merging, thresholding, and morphologic operations; locate the cortical landmarks from the segmented planes. The algorithm has been validated quantitatively with 20 T2W MR images data sets. The mean errors of the Talairach cortical landmarks were below 1.00 mm. It took about 8 seconds for identifying them on P4 3.0 GHz. This fast, robust algorithm is potentially useful in clinic and for research.