Gaussian Markov Random Fields-Based Features for Volumetric Texture Segmentation

A new method based on three dimensional Gaussian Markov Random fields (3D-GMRF) is proposed in this paper for volumetric texture segmentation (VTS). A feature vector is extracted for each voxel in a given volumetric texture image. These feature vectors that consist of the estimated parameters of the GMRF and form the parameter volume are employed to segment volumetric textures. To overcome the issues related to boundaries and isolated voxels, a solution is proposed by sliding an averaging volume inside the parameter volume to assign each voxel a new feature vector derived as the mean of the surrounding voxels that are collected by the averaging volume. Our proposed method is evaluated on a synthetic volumetric texture and compared with another method demonstrating good segmentation performance. A further evaluation is carried out to examine the performance of the method proposed here in the presence of noise to show robustness to noise.