MAP image reconstruction methods using adaptive parameter selection

The performance of an adaptive MAP image reconstruction technique was analyzed. The MAP reconstruction was performed on computer simulated single photon emission computed tomography (SPECT) images. In the adaptive MAP reconstruction, the Gibbs parameter was spatially varied across the image, based on edge features. The performance was compared with conventional MAP techniques which use a constant /spl beta/ for the entire image. The normalized mean square error (NMSE) in reconstruction at the end of 150 iterations was compared for the adaptive and conventional MAP techniques. For the images used in this study, the NMSE for adaptive MAP was lower than NMSE of the conventional MAP.<
>