MAP EM image reconstruction by k-nearest neighborhood selection in emission CT

Maximum A Posteriori estimation using Expectation Maximization algorithm (MAP EM) image reconstruction which was proposed by Green (1990) and Lange (1990) utilized the energy function of some fixed-size neighborhood. Therefore the edge property of a reconstructed image can not be preserved. In this paper, we propose the MAP EM image reconstruction using k-nearest neighbor method to improve the edge property. The results of simulation using the Shepp phantom showed that the method yielded higher quality images than the conventional method in the criteria of the mean square error. This is mainly due to the reproducibility of edge areas in the reconstructed image. (author)