compressive sensing (CS) related research
CS issues in tomography
articles
conference contributions
CS issues in tomography:
articles:
digital breast tomosynthesis
A CS application to clinical data for digital breast tomosynthesis (DBT).
few-view, circular cone-beam (available upon request)
Explains, in quite some detail, the ASD-POCS (adaptive steepest descent-projection onto convex sets) algorithm.
Although the data are computer-simulated projections of the FORBILD jaw phantom (phantom group at Erlangen),
this work performs image-reconstruction for cone-beam computed tomography (CT) systems with realistic dimensions.
A big constraint on CS algorithms for computed tomography is the enormous size of the reconstructed images; this
paper dealt with images containing 4003 voxels.
Another important issue, addressed here, is that the data are generated from a continuous jaw model,
but the image representation is discrete (a voxel array). This
mismatch can create trouble for the application of CS methods.
A slice-image comparison of CS-based image-reconstruction, left column, and a standard iterative algorithm, right column.
The tumor, modeled as a low-contrast sphere, is only visible on the left.
diffraction tomography (available upon request)
Application of CS to wave imaging, using simulated data.
Diagram of diffraction tomography data model
fan-beam image-reconstruction
Our first attempt at CS in CT, using simulated data with the Shepp-Logan phantom.
The algorithm is primitive, but at least it explains the system model for CT in detail.
fan-beam configuration of simulated CT-scanner
conference contributions:
main