Recursive deconvolution of multisensor imagery using finite mixture distributions

Alexia Giannoula, Dimitrios Hatzinakos

Abstract

In this paper, the case where multiple degraded (blurred and noisy) acquisitions of the same scene are available, is investigated. An efficient iterative deconvolution system is introduced, where the multisensor images are fused, based on the classification of each image in a predetermined number of classes that represent the components of a finite mixture of normal densities (FMN). The EM algorithm is utilized for the learning of the FMN model. The recursive employment of the classification and fusion processes, followed by an optimized adaptive filtering, converges to a global enhanced version of the original scene in only a few iterations. Experimental results establish the efficiency of the proposed scheme.
BibTeX
@inproceedings{giannoula2005recursive,
  author = {Alexia Giannoula and Dimitrios Hatzinakos},
  title = {{Recursive deconvolution of multisensor imagery using finite mixture distributions}},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
  year = {2005},
  doi = {10.1109/icassp.2005.1415485},
}

Loading related content...

Figure