Due to intensity inhomogeneities, partial volume effects, as well as organ shape variations, automatic segmentation of abdominal organs has always been a high challenging task. To conquer these difficulties, we employ a pre-labeled atlas (VIP-Man) to supplement anatomical knowledge to the segmentation process. First, an atlas-subject registration is carried out to establish the proper correspondence between the atlas and the subject. The registration consists of two steps. In the global registration step, a similarity transformation is found to eliminate the stature differences. In the organ registration step, organs of interest are registered respectively to achieve better alignments. Second, we utilize the fuzzy connectedness framework to segment abdominal organs of interest from the subject image. Under the guidance of the registered atlas, the seeds and intensity parameters of organs are specified in an auto-adaptive way. Further more, the anatomical knowledge contained in the atlas is blended into the frame work, to make the segmentation result more reliable. To remove possible jags on boundary, a level set smooth method which implements fuzzy connectedness as external speed forces, is utilized on the segmentation result. Our purpose is to accomplish the segmentation task like how anatomy experts do. So far, this approach has been applied to segment organs, including liver, spleen and kidneys, in the female MRI T1 data set from the VHP. All organs of interest are identified correctly, and delineated with considerable precision.