Noise and noise texture in CT images before and after post-processing

Two homogeneous water phantoms were recorded under various CT scanner settings, and the reducing effect of various convolution filters on the noise level was evaluated. The change in noise texture introduced by the filtering process was assessed for a limited range of scanner settings by noise analysis of the autocovariance matrix before and after filtering. The results indicate that reduction of noise expressed as RMSD by a factor of about 8 is possible with the filters available in the software package of the Ohio-Nuclear Delta 50 FS tomograph, with some sacrifice in spatial resolution. Results of preliminary experiments with an inhomogeneous phantom are in agreement with this finding and indicate that filtering allows smaller lesions to be detected. (Auth.)