Prediction of skin quality properties by different multivariate image analysis methodologies

Prediction methods on RGB images have been developed using different feature extraction-based methodologies: Several methodologies based on the use and combination of different wavelet techniques with first and second order statistics have been analysed, as well as a simple SVD decomposition of the images. These different sets of features have been used as sample descriptors to relate the images to an 11 grade skin quality measure using multivariate statistical projection models. These methodologies are employed in a colour-texture integrating scheme, in order to combine both spectral and spatial types of information. The results obtained have been used to compare the methodologies and to confirm the usefulness of MIA for prediction purposes.