Subsampled Blind Deconvolution via Nuclear Norm Minimization
Loading...
Date
2020-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
The Ohio State University
Abstract
Many phenomena can be modeled as systems that preform convolution, including negative effects on data
like translation/motion blurs. Blind Deconvolution (BD) is a process used to reverse the negative effects
of a system by effectively undoing the convolution. Not only can the signal be recovered, but the impulse
response can as well. "Blind" signifies that there is incomplete knowledge of the impulse responses of an
LTI system. Solutions exist for preforming BD but they assume data is fully sampled. In this project we
start from an existing method [1] for BD then extend to the subsampled case. We show that this new
formulation works under similar assumptions. Current results are empirical, but current and future work
focuses providing theoretical guarantees for this algorithm.
Description
Keywords
Blind Deconvolution, Image Reconstruction, Optimization, Nuclear Norm