Supervised segmentation of volume textures using 3D probabilistic relaxation

Authors:
Matthew Deighton;Maria Petrou
Affiliations:
School of Electronics and Physical Sciences, University of Surrey, Guildford, UK;School of Electronics and Physical Sciences, University of Surrey, Guildford, UK and The Institute of Telematics and Informatics, EKETA, Thermi, Thessaloniki, Greece
Venue:
SCIA'03 Proceedings of the 13th Scandinavian conference on Image analysis
Year:
2003

Citing 3
Cited 0

Segmentation of Color Textures

IEEE Transactions on Pattern Analysis and Machine Intelligence
3-D Image Processing Algorithms

3-D Image Processing Algorithms
Texture anisotropy in 3-D images

IEEE Transactions on Image Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

An ieerative 3D probabilistic relaxation scheme has been developed for assigning labels to voxels based on the probabilities that the voxel belongs to each one of a number of known classes. The approach takes account of the probabilities of the neighbouring voxels belonging to each class and of the likely configurations of those labels within the neighbourhood. We apply the approach to the supervised segmentation of a seismic volume, in the example, the probability that a voxel belongs to each class is provided by the application of gradient operators and statistical measures. The iterative relaxation scheme then assigns the most appropriate label to each voxel.