Anais do IX SIBGRAPI'96 (1996), 197-204

The Use of MAP Estimation Techniques in the Tomographic Reconstruction of Poisson Noise Corrupted Images

Abstract:

This work presents new methods for the tomographic reconstruction of images with Poisson noise corrupted projections. The reconstruction method is performed by first filtering the noisy projections under the Maximum a Posteriori criterion and subsequently reconstructing the images through conventional filtering-backprojection methods using the ramp filter. The " a priori" knowledge is incorporated by using several densities, including the Gaussian and densities defined on the non-negative real line. These densities were used to denote the fact that the rates of counting on the projections are non-negative quantities. Experimental results, both simulated and real, indicate that, by using the MAP criterion, it is possible to obtain better performance, as compared to conventional methods of reconstruction, with a very small increase in computational effort.

Full Version:

in Acrobat PDF (108 Kb), and in Postscript gzipped (97 Kb)