Total Variation Regularization of Matrix-Valued Images
2007

Regularization Method for Matrix-Valued Images

publication Evidence: moderate

Author Information

Author(s): Oddvar Christiansen, Tin-Man Lee, Johan Lie, Usha Sinha, Tony F. Chan

Primary Institution: University of Bergen, University of California, Los Angeles

Hypothesis

Can we develop a regularization method for diffusion tensor images that preserves their positive definiteness?

Conclusion

The proposed method effectively regularizes matrix-valued images while preserving edges and positive definiteness.

Supporting Evidence

  • The method preserves edges in the data while regularizing the diffusion tensor.
  • Numerical experiments showed good performance on both synthetic and real data.
  • The proposed method maintains the positive definiteness of the diffusion tensor.

Takeaway

This study created a new way to clean up images of the brain that helps keep important details while removing noise.

Methodology

The authors developed a regularization method for diffusion tensor images using total variation techniques and performed numerical experiments on synthetic and real data.

Limitations

The method's performance may vary with different noise levels and regularization parameters.

Participant Demographics

The study included data from a healthy human volunteer.

Digital Object Identifier (DOI)

10.1155/2007/27432

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication