Characterization of Complex Image Spatial Structures Based on Symmetrical Weibull Distribution Model for Texture Pattern Classification

Figure 15

Feature extraction of TPs, D80, D82, and D84, based on GLCM, GWT, and VZ algorithm and WLD methods. (a) Original TPs, D80, D82, and D84, apparently with great visual similarity; (b) GLCM-based texture features; (c) GWT features, which tend to be the same trend in all the forty subbands; (d) histogram-based texture feature representation based on VZ algorithm proposed by Varma and Zisserman [11], with the first 100 textons learned from the training image with MR8 filter banks class by class. PATCH and PATCH-MRF methods are their more recently texture classification methods, which has the similarity processing scheme but the original pixel patches are used and which can achieve the similar results for TP representation. (e) WLD histogram of D80, D82, and D84.

(a)

(b)

(c)

(d)

(e)