Texture discrimination using brushlet features
Chibiao Chen, Jun Liu, Kap Luk Chan · 2002
We make use of a new adaptive basis of functions, brushlets, which is reasonably well localized with only one peak in frequency. Since a brushlet is a complex valued function, the orientation information is embedded in its phase and can be used for directional texture analysis. A texture classification system is developed using the brushlet representation of original images. Our experiments on a texture database for MPEG-7 core experiments indicate the brushlet features provide a very good texture classification accuracy.