(Results extracted from Dan Yu's dissertation)
Level-set approaches for image segmentation
have recently been proven to be very powerful
due to their intrinsic ability
to accommodate topological variations,
as illustrated below.
As part of his dissertation,
Dan Yu integrated the level-set approach
to contour extraction
into a statistical tomographic image reconstruction method.
This approach is well suited
to image reconstruction problems
with piecewise smooth image characteristics,
such as PET attenuation maps,
as illustrated below.
Furthermore,
the extracted boundary information enables
reconstruction of reasonable attenuation maps
even from low-count transmission scans,
in constrast to conventional reconstruction algorithms
that degrade more rapidly
as the SNR decreases.
Back to Fessler home page