In our development, segmenting the spine shape medical images using efficient techniques. Using MRF(Markov Random Field) clustering to be obtain the global data and shape coherence in manifold spaces of spine models. The different shapes of the input images were previously saved and the shapes can be matched with the input images that which shape is perfect and then finally the segmentation can be employed. The modeling of 3D shape and matching of the shape in the image in 3D were not possible in Matlab. The 2D images can be used as input and clustering is employed using Markov Random Field Model. Markovian models of images may help us to make better image restoration, enhancement or segmentation. However, using segmentation methods based on Markov Random Field (MRF) models requires a huge computing power and quite a lot of time. For these reasons, MRF methods are usually used in o€-line tasks, never in a real-time processing environment. Recently multiprocessor arrays, transputers or distributed computer systems have been used for MRF processing, but all of them are power-consuming and still slow considering the standard video-rate.