Blind Image Deconvolution
Muhammad Hanif
NICTA SML SEMINARDATE: 2012-10-25
TIME: 11:15:00 - 12:00:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
Blind image deconvolution (BID) is a long standing research problem in image processing community, with application in remote sensing, medical imaging, astronomy, photography, and more. Classical image restoration aims at recovering an estimate of original image, assuming the degradation completely known. However, this assumption is not always satisfied. In most practical imaging situations it is often costly or physically impossible to obtained prior information about the degradation process. This motivates and direct us to blind image restoration, where the primary goal is to recover an estimate of original image given little or no prior knowledge about the degradation process. This is a difficult and ill-posed problem as the uniqueness and stability of the solution is not guaranteed. We approached BID problem using Bayesian frame work, with wavelet based approaches. Specifically, the application of Expectation Maximization (EM) and Kullback-Leibler Divergence algorithms, along with the consideration of natural image statistics can result in more efficient methods.
BIO:
