Figure 1.
The Statistical_MR8 filter bank consists of a series of anisotropic filters (an edge and a bar filter at 6 orientations and 3 scales), and 2 rotationally symmetric ones (a Gaussian and a Laplacian of Gaussian) [16]–[17].
Figure 2.
Illustration of computation for Statistical_Fractal [20].
Figure 3.
Illustration of texton feature for Statistical_Joint [18]–[19].
A 3*3 image patch is converted to a 1*9 texton through recording intensity values row by row.
Figure 4.
Major framework of texton feature extraction.
a) statistical texton; b) binary texton.
Figure 5.
The top row shows 3 texture images.
The central image patch (highlighted by red rectangle) is matched with an edge filter at all orientations. The magnitude (absolute value) of the filter response versus the orientation is plotted in the bottom row.
Figure 6.
The Statistical_MR8 filter banks are divided equally into two groups.
Figure 7.
An illustration example of Statistical_MR8 and Binary_MR8.
Figure 8.
A 7*7 patch is divided into 6 blocks.
Table 1.
45 kinds of sub-textons with 7 bits and their labels.
Figure 9.
An illustration example of Statistical_Joint and Binary_Joint.
Figure 10.
Textures from the Columbia-Utrecht database.
In this work, all images are converted to monochrome, so colour is not used to discriminate between different textures.
Table 2.
Classification rate (%, mean±standard deviation) for CUReT database.
Table 3.
Classification rate (%, mean±standard deviation) for Statistical_MR8 and Statistical_Joint on CUReT database with different training conditions.
Figure 11.
These two patches may generate the same statistical texons, but different binary textons.
Table 4.
Classification rate (%, mean±standard deviation) for noise sets.
Figure 12.
Samples of the 25 textures in UIUC database.
Table 5.
Classification rate (%, mean±standard deviation) for UIUC database.
Table 6.
Classification rate (%, mean±standard deviation) for different resolution.
Figure 13.
Samples of the 10 textures in KTH-TIPS database.
Table 7.
Classification rate (%, mean±standard deviation) for KTH-TIPS database.
Table 8.
Average time (seconds) cost on feature extraction for different databases.