Abstract:

In order to process a 3D image by the stochastic method based on 2D Markov Random Field(MRF), the authors proposed a 3D MRF image model. The authors extend neighbor, clique and potential function from planar domain to 3D space, define 3D random field, 3D neighbor and 3D clique, point out the position of neighbor and clique in 3D space and establish the practical expression of the potential function. The authors also explain the space interaction and probability distribution between them and not only solve the problem of 3D modeling theoretically but also propose a practical stochastic tool for 3D image computer process as well.

Key words: 3D image, Mar kov random field model, Gibbs distributions