Mathematics Colloquia and Seminars

Return to Colloquia & Seminar listing

Sparse Image Representations with Deformations

Colloquium

Speaker: Professor Stephane Mallat, Chair of Applied Math., Ecole Polytechnique, Paris
Location: 693 Kerr
Start time: Fri, Jan 21 2000, 4:10PM

Constructing sparse representations often leads to simple and efficient algorithms for applications such as compression, estimation, inverse problems and pattern recognition. This explains the success of new orthogonal bases such as wavelets, which will be illustrated by examples. To improve current results, it is necessary to better take into account physical models and deformations. Modeling and computing these deformations provides representations that are more efficient than direct decompositions in optimized bases. An approach with partial differential equations is presented.