Chair: Jr-Jen Huang, Purdue University, USA


Three Dimensional Segmentation of Optical Microscope Images


M. Razaz, University of East Anglia (U.K.)
J. A. Bangham, University of East Anglia (U.K.)
R. A. Lee, University of East Anglia (U.K.)
R. W. Harvey, University of East Anglia (U.K.)

Volume 1, Page (NA), Paper number 611


Three dimensional optical microscopy images are noisy and blurred, have nonuniform background, and contain objects which do not usually have sharp edges or may have noise induced boundaries. As a result, traditional segmentation techniques are not suitable for this type of applications. We present a novel methodology based on a combination of 3D nonlinear restoration and morphological sieving which can be used to successfully segment 3D optical microscopy images. The nonlinear restoration removes the blur and noise aberrations from such real images and the sieve algorithm segments out their subcellular features. The methodology is discussed and experimental results using both synthetic and real 3D images are presented.

ns970611.pdf (Scanned)

ns970611.pdf (From Postscript)


Analysis of the ML-EM Algorithm for Nonlinear Reconstruction of Positron Emission Tomography Images


F. Boschen, University of Wuppertal (Germany)
A. Kummert, University of Wuppertal (Germany)
H. Herzog, Research Center Juelich KFA (Germany)

Volume 1, Page (NA), Paper number 612


Positron emission tomography (PET) is a technique that has opened new facilities to study the metabolic activity of human body. In the last years many algorithms have been developed for reconstructing tomography images. The oftenly used maximum likelihood expectation maximization algorithm (ML-EM) seems to be a stable method and was developed by Shepp and Vardi [1] in 1982. However, the ML-EM algorithm causes some serious problems in the context of the application considered. It is an iterative procedure and converges to a stationary point, however, the reconstructed image seems to be distorted by superposed high frequency noise. In this paper it is shown, that the ML-EM Algorithm is not based on significant statistical properties in our problem, which has been verified by respective investigations. It is shown, that the algorithm converges to an 'optimal' solution in a mathematical sense. The convergence characteristics of the algorithm are discussed by means of examples.

ns970612.pdf (From Postscript)