Abstract
In micro-CT imaging, structures that are small w.r.t the image resolution
are hard to segment because of partial volume effects.
In this paper, we show that if the object consists of
a small number of different materials, each having a constant
density, the resolution of the tomographic reconstructions
can be dramatically improved by using prior information
on the grey values of the scanned objects, resulting in
much more accurate segmentations. The proposed method is
based on an upsampling of the reconstruction grid, combined
with the discrete algebraic reconstruction technique (DART)
[1], in which the scanned object is assumed to be composed of
homogeneous materials. Experiments on simulated CT data
of foams show that the proposed method indeed generates
significantly better segmentations compared to conventional
methods.
are hard to segment because of partial volume effects.
In this paper, we show that if the object consists of
a small number of different materials, each having a constant
density, the resolution of the tomographic reconstructions
can be dramatically improved by using prior information
on the grey values of the scanned objects, resulting in
much more accurate segmentations. The proposed method is
based on an upsampling of the reconstruction grid, combined
with the discrete algebraic reconstruction technique (DART)
[1], in which the scanned object is assumed to be composed of
homogeneous materials. Experiments on simulated CT data
of foams show that the proposed method indeed generates
significantly better segmentations compared to conventional
methods.
| Original language | English |
|---|---|
| Title of host publication | First international conference on image formation in X-ray computed tomography |
| Publication status | Published - Jun 2010 |
| Event | Unknown - Duration: 1 Jun 2010 → … |
Conference
| Conference | Unknown |
|---|---|
| Period | 1/06/10 → … |
Keywords
- micro CT
- super resolution
- metal foams
- discrete tomography