9
as resizing only changes the dimensions of the image and
does not capture the same level of detail that would be pres-
ent in the original scene. It can mimic the visual appearance
to some extent, but it might not fully replicate the effects
of perspective, focus, and depth that result from changing
the physical distance between the camera and the subject.
However, it is a helpful tool to study the visual impact of
distance changes.
Various resolutions, namely 1, 2, and 4 pixels/mm (25,
50 and 100 pixels/in) were used in this study. The resolu-
tion of 1 pixel/mm was the recommended minimum reso-
lution in the previous study (Yu et al. 1997)methods for its
characterisation and analysis are poorly developed. Banding
texture was obtained manually from the coal face and core
at a minimum resolution of 1 mm. Window filtering was
used to determine the optimum resolution (30–50 mm,
and the other two resolutions were higher than the recom-
mended minimum value. In addition, when the GLCM
parameters are compared with different resolutions, there
are two ways to make the comparison. The first one is
based on the patch size with the same pixel numbers, and
different covered areas can be obtained because of the dif-
ferent resolutions. For this method, a patch size of 50×50
pixels was used for the analyses, leading to patch sizes of
50×50 mm, 25×25 mm, and 12.5×12.5 mm, respectively.
The second method is based on the patch size with the same
covered areas, and different pixel numbers are required
when the resolution varies. For this method, a patch size
of 25×25 mm was used for the analyses, resulting in patch
sizes of 25×25 pixels, 50×50 pixels, and 100×100 pixels,
respectively. The calculated image features with varying res-
olutions and two different comparison methods are shown
in Figure 7.
For the first method, the covered area reduces with the
increasing resolution. Although the increasing resolution
helps to capture finer details, the decreasing area increases
the randomness of the features, potentially leading to
higher variation. These conflicting influences may show
different effects on different features. Figure 7 (a) shows
that the variation in contrast, correlation, and dissimilarity
decreases with increasing resolution, while the variation in
energy, homogeneity, deviation and mean becomes larger
with the increase in resolution. For the second method, the
same area is covered, and the higher resolution can cap-
ture finer details without losing the texture information at
a coarser spatial scale. It helps to capture consistent image
features. It can be found from Figure 7 (b) that there is a
general trend of reducing variation with increasing resolu-
tion. The second method can provide better comparable
results than the first one. Thus, it is recommended to use
the second method for the comparison with different res-
olutions. The less variation with increasing resolution for
the second approach indicates that the data are more con-
centrated. Since the features calculated for BC and BBC
(a) Box plot of various GLCM features of same patch size in pixel.
(b) Box plot of various GLCM features of same covered area.
Figure 7. Comparison of GLCM features with varying resolutions
as resizing only changes the dimensions of the image and
does not capture the same level of detail that would be pres-
ent in the original scene. It can mimic the visual appearance
to some extent, but it might not fully replicate the effects
of perspective, focus, and depth that result from changing
the physical distance between the camera and the subject.
However, it is a helpful tool to study the visual impact of
distance changes.
Various resolutions, namely 1, 2, and 4 pixels/mm (25,
50 and 100 pixels/in) were used in this study. The resolu-
tion of 1 pixel/mm was the recommended minimum reso-
lution in the previous study (Yu et al. 1997)methods for its
characterisation and analysis are poorly developed. Banding
texture was obtained manually from the coal face and core
at a minimum resolution of 1 mm. Window filtering was
used to determine the optimum resolution (30–50 mm,
and the other two resolutions were higher than the recom-
mended minimum value. In addition, when the GLCM
parameters are compared with different resolutions, there
are two ways to make the comparison. The first one is
based on the patch size with the same pixel numbers, and
different covered areas can be obtained because of the dif-
ferent resolutions. For this method, a patch size of 50×50
pixels was used for the analyses, leading to patch sizes of
50×50 mm, 25×25 mm, and 12.5×12.5 mm, respectively.
The second method is based on the patch size with the same
covered areas, and different pixel numbers are required
when the resolution varies. For this method, a patch size
of 25×25 mm was used for the analyses, resulting in patch
sizes of 25×25 pixels, 50×50 pixels, and 100×100 pixels,
respectively. The calculated image features with varying res-
olutions and two different comparison methods are shown
in Figure 7.
For the first method, the covered area reduces with the
increasing resolution. Although the increasing resolution
helps to capture finer details, the decreasing area increases
the randomness of the features, potentially leading to
higher variation. These conflicting influences may show
different effects on different features. Figure 7 (a) shows
that the variation in contrast, correlation, and dissimilarity
decreases with increasing resolution, while the variation in
energy, homogeneity, deviation and mean becomes larger
with the increase in resolution. For the second method, the
same area is covered, and the higher resolution can cap-
ture finer details without losing the texture information at
a coarser spatial scale. It helps to capture consistent image
features. It can be found from Figure 7 (b) that there is a
general trend of reducing variation with increasing resolu-
tion. The second method can provide better comparable
results than the first one. Thus, it is recommended to use
the second method for the comparison with different res-
olutions. The less variation with increasing resolution for
the second approach indicates that the data are more con-
centrated. Since the features calculated for BC and BBC
(a) Box plot of various GLCM features of same patch size in pixel.
(b) Box plot of various GLCM features of same covered area.
Figure 7. Comparison of GLCM features with varying resolutions