Welcome to SIAM at Tufts!
The Tufts chapter of the Society for Industrial and Applied Mathematics promotes exploration in applied mathematics for the student communities at Tufts.
Meetings are held every other Wednesday from 12:00pm to 1:15pm in either the Clarkson Conference Room or Classroom 101 in Bromfield-Pearson, unless otherwise noted.
March 4, 2015
Speakers: Kye Taylor, PhD, Mathematics Department, Tufts University
Title: Reordering of image pixels using patches to achieve image compression
Abstract: Recent improvements in computing demand the processing and analysis of very large datasets in the form of matrices in a variety of fields. Often the analysis requires the creation of low-rank approximations to the matrices for the purpose of compression. In this talk, I will present a new approach to compressing images. This approach identifies an image as a series of non-overlapping contiguous patches, and compresses a reordered version of the image matrix based on these patches. Low-rank approximations can then be utilized once a reordering of the pixel entries of the matrix is performed. I will demonstrate the approach when using the Singular Value Decomposition (SVD) and Nonnegative Matrix Factorization (NMF) low-rank approximations. I will also demonstrate that the approach produces better quality image reconstructions using less memory for a smaller computational cost than when compressing without reordering.