Matlab version

The Matlab scripts that can be downloaded here below have been tested with Matlab 2007/a.


The functions used to extract the colour texture features can be downloaded here below. Classification is based on the nearest-neighbour rule (1-NN) with L1 distance. Implementations of this classifer are available in many toolboxes and packages, such as: Classification Toolbox for Matlab, PRTools and WEKA.

Class CTD:ColourTextureDescriptor
Normalized colour space representation
Average colour differences
Tex-Mex colour features

Class CTD:PurelySpectral
Colour histogram
Fuzzy colour histogram

Class CTD:PurelySpectralMarginalization2D
Chromaticity moments

Class CTD:PurelySpectralMarginalization1D
Three marginal histograms (sRGB)
Three marginal histograms (rgb)
Three marginal histograms (I1I2I3)
One marginal histogram (H)
One marginal histogram (V)
Two marginal histograms (H+V)
Colour statistics (mean, sRGB)
Colour statistics (mean, rgb)
Colour statistics (mean + std + 3-5th, sRGB)
Colour statistics (mean + std + 3-5th, Lab)
Colour statistics (percentiles, sRGB)

Class CTD:ColourIndexing
Co-occurrence matrices
Multilayer CCR

Class CTD:SpectralMarginalizationIntraChannel
Gabor features
Gabor features on Gaussian colour model
Wavelets and co-occurrence histograms

Class CTD:SpectralMarginalizationInterChannel
Histograms of colour ratios

Class CTD:SpectralMarginalizationIntraAndInterChannel
Opponent Gabor features
Opponent Colour LBP
Integrative co-occurrence matrices
Colour ranklets

Class CTD:DisjointSpectralAndGrayscale
Gabor and chromatic features
DT-CWT and chromatic features
Co-occurrence matrices and chromatic features
Co-occurrence matrices and colour percentiles
LBP and colour percentiles


Most of the functions listed above depend on other functions that are available here: In order to correctly run the code listed above, proceed as follows:

1) create the directory structure: code/dependencies;
2) download the above functions in code;
3) download and unzip the file in dependencies;
4) go to Matlab -> File -> Set Path -> Add with Subfolders and add the directory code.


Gabor features are based on the SIMPLEGABOR toolbox for Matlab developed at the Lappeenranta University of Technology, Finland.

Dual-Tree Complex Wavelet Features are based on the Matlab implementation of Wavelet Transforms available at the Polytechnic Institute of New York University, USA