Multi-resolution diffusion tensor filter for preserving noise power spectrum in low-dose CT imaging

Recent emphasis on reducing radiation dose associated with CT imaging has highlighted the need for methods to mitigate the effects of increased image noise without compromising spatial resolution. We present a novel 3D multiresolution non-linear anisotropic diffusion image processing method, which incorporates anisotropic diffusion tensors computed at multiple spatial scales based on Laplace pyramid decomposition of the reconstructed volume image. Diffusion processing is applied independently to the component images prior to the Laplace reconstruction. Low-dose anatomic and phantom volume images were processed and compared to a previously described non-linear diffusion algorithm based on an adaptive scalar diffusion parameter. The new method provides better edge preservation at the same noise reduction level. The noise power spectra displayed a substantial reduction in low-frequency noise while maintaining some degree of higher frequency noise component, suggesting an improvement in detectability of low contrast objects and fine structures.