MLEM Image Reconstruction Algorithm Based on Physical Constraints

Maximum likelihood expectation maximization (MLEM) algorithm is an iterative method used to reconstruct radiation image in coded aperture gamma camera. This method works better in restraining noise when compared with traditional linear algorithms, but its convergence rate is low. Furthermore, noise in the reconstructed image will increase sharply for the excessive iterations when projection image contains too much noise. For the above mentioned reasons, it is necessary to optimize the MLEM algorithm and improve its performances. In this paper, physical constraints were added to the MLEM iteration based on direct demodulation method, and image data restored by the cross correlation method were used as the initial value of the new iteration. Simulation result shows that convergence rate of the iteration is accelerated and the convergence of the reconstructed image is better using the physical constraints during the iteration process. The method is effective to improve the quality of reconstructed image.