The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B3b. Beijing 2008 
Table 1. Discriminant analysis results for binary images under 
the Gliding-Box algorithm 
A better distinction between images As and Bs is generated 
when the same Gliding-Box algorithm is applied to grayscale 
images (table 2). In this case, 80% of the image samples were 
correctly classified. So, we can conclude that the higher the 
radiometric resolution of an image, the better will be its texture 
discrimination and therefore a better socioeconomic pattern 
distinction of image samples. 
In this section we make a brief discussion about the lacunarity 
results obtained with the two algorithms presented before, 
considering their possibilities in the discrimination of textures 
from the image samples selected in the last section. In order to 
permit a better comprehension of the difference between these 
algorithms, the mean lacunarity, Lm, of the selected box sizes 
was calculated and compared through a statistical discriminant 
analysis. 
Table 1 shows the discrimiant analysis results from the mean 
lacunarity values of binary images As and Bs under the 
Gliding-Box algorithm. As we can see, only 50% of these 
images were correctly classified (images not shaded in Table 1), 
that is, the previous group defined by the inhabitability 
conditions of the urban area where the image was extracted, is 
the same as the posterior group generated through the 
discriminant analysis of the mean lacunarity value of this 
particular image. 
Image 
Lm 
Score 
Group 
Previous 
Posterior 
Al 
1.29 
1.21 
1 
1 
A2 
1.25 
0.20 
1 
1 
A3 
1.22 
-0.27 
1 
2 
A4 
1.23 
-0.09 
1 
2 
A5 
1.28 
0.85 
1 
1 
A6 
1.29 
11.9 
1 
1 
A7 
1.22 
-0.25 
1 
2 
A8 
1.23 
-0.16 
1 
2 
A9 
1.22 
-0.35 
1 
2 
AIO 
1.23 
-0.24 
1 
2 
All 
1.24 
0.18 
1 
1 
A12 
1.23 
-0.13 
1 
2 
A13 
1.21 
-0.50 
1 
2 
A14 
1.23 
-0.23 
1 
2 
A15 
1.29 
11.06 
1 
1 
B1 
1.22 
-0.40 
2 
2 
B2 
1.22 
-0.41 
2 
2 
B3 
1.24 
0.01 
2 
1 
B4 
1.23 
-0.13 
2 
2 
B5 
1.27 
0.77 
2 
1 
B6 
1.19 
-10.38 
2 
2 
B7 
1.20 
-0.88 
2 
2 
B8 
1.23 
-0.20 
2 
2 
B9 
1.19 
-0.97 
2 
2 
BIO 
1.26 
0.53 
2 
1 
Bll 
1.25 
0.22 
2 
1 
B12 
1.21 
-0.57 
2 
2 
B13 
1.26 
0.62 
2 
1 
B14 
1.24 
0.01 
2 
1 
B15 
1.23 
-0.13 
2 
2 
Image 
Lm 
Score 
Group 
Previous 
Posterior 
Al 
1.15 
0.03 
1 
1 
A2 
1.18 
0.67 
1 
1 
A3 
1.14 
-0.03 
1 
2 
A4 
1.14 
-0.03 
1 
2 
A5 
1.18 
0.63 
1 
1 
A6 
1.18 
0.68 
1 
1 
A7 
1.14 
-0.10 
1 
2 
A8 
1.17 
0.49 
1 
1 
A9 
1.18 
0.66 
1 
1 
AIO 
1.16 
0.20 
1 
1 
All 
1.15 
0.10 
1 
1 
A12 
1.16 
0.26 
1 
1 
A13 
1.16 
0.22 
1 
1 
A14 
1.15 
0.00 
1 
2 
A15 
1.18 
0.64 
1 
1 
B1 
1.10 
-0.96 
2 
2 
B2 
1.10 
-0.83 
2 
2 
B3 
1.11 
-0.74 
2 
2 
B4 
1.11 
-0.77 
2 
2 
B5 
1.16 
0.30 
2 
1 
B6 
1.12 
-0.54 
2 
2 
B7 
1.14 
-0.02 
2 
2 
B8 
1.14 
-0.03 
2 
2 
B9 
1.11 
-0.63 
2 
2 
B10 
1.12 
-0.55 
2 
2 
Bll 
1.12 
-0.49 
2 
2 
B12 
1.31 
32.47 
2 
1 
B13 
1.11 
-0.76 
2 
2 
B14 
1.10 
-0.86 
2 
2 
B15 
1.11 
-0.77 
2 
2 
Table 2. Discriminant analysis results for grayscale images 
under the Gliding-Box algorithm 
Table 3 shows the discriminant analysis results from grayscale 
images under the Differential Box-Counting algorithm. In this 
case, 90% of the selected images were correctly classified. Only 
2 image samples (B5 and B7) were assigned to different groups. 
Considering these results, we can conclude that the DBC 
applied in grayscale images is the most appropriate algorithm to 
discriminate texture from urban areas with different 
inhabitability conditions. 
420 
BMHM