Regularized Reconstruction of Vector Fields

(Click to enlarge.)

Description

In the following references we presented a general framework for scale- and rotation-invariant regularized denoising and reconstruction of vector fields, and considered the important examples, namely curl-divergence regularization with L1 and L2 norms. The present package provides MATLAB implementations of these two schemes, using the iterative reweighted least squares (IRLS) method.

Additional code for parameter optimization using and oracle or estimates of the SNR using Stein's Unbiased Risk Estimate (SURE) and Generalized Cross-Validation (GCR) (under the Gaussian noise hypothesis) is also provided.

See the README file for more information.

References

P. D. Tafti and M. Unser, On regularized reconstruction of vector fields, IEEE Trans. Image Process., vol. 20, no. 11, pp. 3163–78, 2011.
P. D. Tafti, R. Delgado-Gonzalo, A. F. Stalder, and M. Unser, Variational enhancement and denoising of flow field images, Proc. 8th IEEE Int. Symp. Biomed. Imaging (ISBI 2011), pp. 1061–4, Chicago, IL, 2011.

Licence

See the general conditions of use at the bottom of this page.

The present code is provided freely for non-commercial and educational use, but without any guarantees of any kind. Use at your own risk. :-)

If you use this code or derivatives of it, or results obtained using the code or its derivatives in publications, we would appreciate references to the above two papers (the first being more important). In addition, we request that you do not re-distribute this code or derivatives of it without our consent and permission.

Author:
Pouya Dehghani Tafti (pouya dot tafti at a3 dot epfl dot ch)

Dates:
09 Feb. 2012 (current release)
?? Feb. 2011 (this implementation)

Regularized Reconstruction of Vector Fields

Description

References

Licence

Download