A parametric and indexing method for faces and textures recognition

Image recognition which is one large library of images is really an important topic in multimedia research. Several methods such as wavelet transform or raw pyramid transform methods are used in image recognition. Many other techniques which are based on statistical images analysis exist. These methods using statistics give interesting results without being touching the images characteristics. The proposed approach is based on multi resolution and wavelet transform. The chosen distance for distance comparison is the Kullback-Leiber one. This approach is tested for texture images and for facial recognition one.