R. Lakmann and L. Priese,
A Reduced Covariance Color Texture Model for
Micro-Textures,
In The 10th Scandinavian Conference on Image Analysis,
Lappeenranta, Finland, pp. 947-953, June 1997.
R. Lakmann and L. Priese,
A Reduced Covariance Color Texture Model for
Micro-Textures,
In The 10th Scandinavian Conference on Image Analysis,
Lappeenranta, Finland, pp. 947-953, June 1997.Color and texture, two main features in image analysis, are integrated into a unique color texture model. A reduced color covariance texture model is described which is based on second order statistics. In this texture model natural color micro-textures are represented by color histograms and color covariance matrices. While the color histograms store the color distribution of the image a set of covariance matrices represents the spatial interrelation of color pixels. The significance of the parameter set is tested by the results of a re-synthesis algorithm. The histogram and covariance parameters of natural color textures are calculated and taken as an input for the re-synthesis of the same texture. Then the synthesis result is compared with the original. A high similarity indicates an adequate significance of the parameter set. Various experiments are performed to find a minimal set of significant parameters for the covariance texture model. The histogram quantisation, the size, number and density of the covariance matrices are reduced to define a minimal parameter set for color micro textures.