A robust Estimation Algorithm for Emission Tomography
CG Zhejiang · Signal Processing · 1996
In this paper, the authors investigate the robustness of the EM (Expectation-Maximum)algorithm for Emission Tomography（ECT）. According to the central limit theorm, the Possion distribution will approximate to the Gaussion distribution when the intensity (or expectation) of the Possion is enough large. So, we convert the Possion model of Emission Tomography into Gaussion and find that the MLE (maximum likelihood estimation)of the problem isn't robust. We use the M- estimation (a robust estimation) to improve the commonly-used MLE algorithm (EM,for example)and obtain a robust EM (REM) algorithm.The computer simulation demonstrate to be satisfactory.