Colour Spectral Analysis for Spatial Structure Characterization of Textures in IHLS Colour Space
Résumé
In this article, linear prediction model based approach for colour texture characterization and classification in the Improved Hue Luminance and Saturation colour space is presented. Pure chrominance structure information is used in addition with normally used luminance structure information for colour texture classification. Hue and saturation channels of a colour image in IHLS colour space are combined through a complex exponential to give a single channel which holds all the chrominance information of the image. Two dimensional complex multichannel versions of Non-Symmetric Half Plane Autoregressive model, Quarter Plane Autoregressive model and Gauss Markov Random Field model are used to perform parametric power spectrum estimation of both luminance and the \textquotedblleft combined chrominance\textquotedblright\ channels of the image. Accuracy and precision of these spectral estimates are proven quantitatively by performing tests on a large number of images. Spectral distance measures are calculated for the spectral information of luminance and chrominance channels individually as well as combined through a combination coefficient. Using these distance measures, colour texture classification is done with $k$-nearest neighbour algorithm. A comparison of experimental results in IHLS with the ones in RGB indicate the significance of using IHLS for such analysis. They also show that colour texture characterization and percentage classification obtained by combined luminance and chrominance structure information is better than the colour texture classification done using only the luminance structure information.