Segmentation of Random textures

This page is related to the linked paper.
We propose a combination of linear and morphological approach for the characterization and segmentation of binary and digital random textures.
Images are available to the community to compare with different approaches.

Download images

The final misclassified pixel rate we achieve are :
 

Error
Fig. LDA
SVM
2.a
15.7 %
9.7 %
2.b
14.8 %
9.4 %
2.c
13.5 %
12.1 %
2.d
18.1 %
19.3 %


                 (2a)                                               (2b)                                             (2c)                                             (2d)


          Mask

Contact : Aurelien.Cord(at)lcpc.fr
[Accueil ] [LIVIC]