WassMap: A Wasserstein Distance-Based Dimensionality Reduction Technique for Image Data
Sparsity in Imaging
WassMap: A Wasserstein Distance-Based Dimensionality Reduction Technique for Image Data
Series: Sparsity in Imaging
Location: Department of Electrical & Computer Engineering ECE 530
Presenter: Nick Henscheid, Department of Mathematics, University of Arizona
Manifold models for image data can be used to perform a wide array of analysis and restoration tasks. Typically, such models make various assumptions about the underlying structure of the data, one the most common being an assumption about the metric structure of the data. We'll discuss a new dimensionality reduction technique that assumes a Wasserstein metric on image data, and compare its performance to existing methods on a host of synthetic test images.
(Please note the venue: ECE 530)