Deconvolution and Singular Value Decomposition.
Abstract
We explore numerical methods for solving multi-dimensional deconvolution and inverse scattering problems, and are particularly interested in techniques based on the singular value decomposition (SVD) to find low rank matrix approximations. Our emphasis will be on both speed and accuracy. We will develop new techniques for numerical multilinear algebra, in particular for computing the SVD of a compound matrix. We will investigate implementation on parallel hardware, and the actual prototyping of such designs. We will also study fault tolerance issues associated with such special-purpose VLSI processors, and develop methods to detect possible ill-conditioning of the underlying deconvolution problem
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 17, 1994
- Accession Number
- ADA286582
Entities
People
- Franklin T. Luk
Organizations
- Rensselaer Polytechnic Institute