Textural features with illumination and rotation invariance


Authors:
Pavel VĂ¡cha, Michal Haindl

Institute of Information Theory and Automation of the ASCR



The presented features for robust recognition of natural and artificial materials (textures) are based on a type of Markov Random Field representation, which is estimated in a very efficient way. The features are invariant to illumination colour, cast shadows, and texture rotation; they are also robust to illumination direction and degradation by Gaussian noise. More details can be found in my PhD thesis or in published articles. The performance of our textural features is presented in the following demonstrations, which include various setups and texture databases.


Applications and experiments:
Mobile Wood Veneer Recognition
(Veneers)
 
Rotation and illumination invariance
(ALOT)
 
Content-Based Tile Retrieval System
(Sanita.cz)
 
Illumination spectrum/colour invariance
(Outex)
 
Illumination direction robustness
(CUReT)