When testing MACE with qGGMRF denoiser, we observed that while the converged result seems correct (in the sense that it is close to MBIR recon result), the convergence is slow.
See the slides below for the experiment details (this is included in Lilly weekly update slides on 02/23/2023).
mace_qGGMRF_slow_convergence.pptx
Per Dr. Buzzard's request, I'm recording this issue here, so that we do not lose track of it.
When testing MACE with qGGMRF denoiser, we observed that while the converged result seems correct (in the sense that it is close to MBIR recon result), the convergence is slow.
See the slides below for the experiment details (this is included in Lilly weekly update slides on 02/23/2023).
mace_qGGMRF_slow_convergence.pptx
Per Dr. Buzzard's request, I'm recording this issue here, so that we do not lose track of it.