Share Email Print

Proceedings Paper

The pixon and Bayesian image reconstruction
Author(s): Richard Charles Puetter; Robert K. Pina
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

The primary purpose of this paper is to present the theory of the pixon, its role in Bayesian image reconstruction, and a new method for selecting a pixon basis using fractal dimension concepts: the Fractal Pixon Basis (FPB). In addition, we review our recent work involving the Maximum Residual Likelihood (MRL) criterion for the goodness of fit (GOF) and the Uniform Pixon Basis (UPB). The MRL GOF statistic is based on the autocorrelation of the residuals and eliminates spatially correlated residuals which commonly occur in image reconstruction methods. The presence of spatially correlated residuals causes photometric inaccuracy. Consequently, photometry is greatly enhanced by using the MRL statistic. We also show that through the use of the UPB image representation, a 'Super-Maximum Entropy' solution can be obtained in which entropy is maximized exactly. We present reconstructions obtained with the above methods and compare them to those obtained with the best Maximum Entropy algorithms and demonstrate that use of UPB/MRL concepts provides consistently superior solutions. Finally, we present reconstructions which demonstrate that the FPB obtains the best results so far, significantly superior to both UPB/MRL and ME-based reconstructions.

Paper Details

Date Published: 20 October 1993
PDF: 12 pages
Proc. SPIE 1946, Infrared Detectors and Instrumentation, (20 October 1993); doi: 10.1117/12.158693
Show Author Affiliations
Richard Charles Puetter, Univ. of California/San Diego (United States)
Robert K. Pina, Univ. of California/San Diego (United States)

Published in SPIE Proceedings Vol. 1946:
Infrared Detectors and Instrumentation
Albert M. Fowler, Editor(s)

© SPIE. Terms of Use
Back to Top