A Novel Local Human Visual Perceptual Texture Description with Key Feature Selection for Texture Classification

Figure 4

Example of generating multiple training and testing texture images from a given texture. The texture is cut into patches with the size of , and then the 64 patches are randomly divided into 8 groups, each of which contains 8 patches. In each group, the 8 patches are randomly combined (with repetition) into 8 rows and 8 columns and then merged into one image. This step is repeated 8 times in each group, resulting in 8 images with the same type of texture. (a) The sample texture. (b) 64 patches of the sample. (c) Regenerated texture by merging the 8 partitions in one group randomly. (d) Regenerated texture by merging the 8 partitions in another group randomly.

(a)

(b)

(c)

(d)