Digital Signal Processing Reference
In-Depth Information
The exterior, cross-sections (X=40, Y=40, Z=40) and pore space of the true image
are shown respectively in Fig. 5 (a), (b) and (c). 0.5% sample points randomly
collected from the true image were used as original conditional data, as shown in Fig.
5(d). The proportion of pore space points in the total points is 0.1742.
(a) (b) (c)
Fig. 4.
The training image. (a) exterior; (b) cross-section(X=40, Y=40, Z=40); (c) pore space
(a) (b) (c) (d)
Fig. 5.
The true image. (a) exterior; (b) cross-section(X=40, Y=40, Z=40); (c) pore space; (d)
sample points
MPS reconstruction of porous media was made by using the training image and
sample points. 3D data templates were used to simulate unsampled nodes. Fig. 6(a),
(b) and (c) are respectively the exterior (80×80×80 voxels), cross-sections (X=40,
Y=40, Z=40) and pore space of the image reconstructed by MPS. It is seen that the
pore space in the horizontal and vertical cross-sections are irregular (shown in Fig.
6(b)), and that the pore space regenerates the long-range connectivity (shown in Fig.
6(c)) existing in the training image. The porosity of the reconstructed image is 0.1743
which is very close to that of the true image.
(a) (b) (c)
Fig. 6.
The MPS-reconstructed image using grid data templates. (a) exterior; (b) cross-section
(X=40, Y=40, Z=40); (c) pore space.
