Conclusions

We developed a new method for accurate analysis and restoration of noisy data. This novel approach, based on ideas from geometry, analysis and statistics, delivers state-of-the-art performance, often outperforming competitive alternatives such as wavelet-based techniques.

Its flexibility allows the use of this anisotropic estimator for several, different, applications.

The current implementation and the shown examples are for 2D signals, but the method itself is general and can be used for higher dimensional data (eg. video or volumetric data).

start again

references

---

Tampere University of Technology - Institute of Signal Processing - Transforms and Spectral Methods Group