Outlier robust statistical methods in image analysis
by Christine Mueller
Abstract: Images contain often sharp edges and corners. Hence for denoising noisy images, it is important that not only the noise is deleted but also edges and corners are preserved. Since outlier robust methods follow the majority of the data, they delete noise and preserve edges. However they do not preserve corners. In this talk, it is shown that only special outlier robust estimators can preserve corners. These are redescending M-estimators with a special initial estimator. These estimators can preserve in principle corners with arbitrary small angle. Redescending M-estimators can be also used to detect geometrical objects in images.
For More Information: Contact: Guoqi Qian firstname.lastname@example.org