Invariant texture segmentation with reduced illumination sensitivity

A method for rotation and scale-invariant texture segmentation is proposed, which can also be employed for object recognition based on pattern analysi s in noisy images. The segmentation scheme is based on a supervised rotatio n and scale-invariant texture recognition using multi-channel polar logarit hmic Gabor filters for feature extraction. The polar logarithmic arrangemen t works like a Fourier-Mellin descriptor providing orientation and scale in variance. The classification of the features is carried out by symmetric ph ase-only matched filtering. The classification accuracy is about 90% at arb itrary rotation angle and for scale factors between 0.25 and 4.0. Rotation angle and scale factor can be determined with high precision by the classif ication scheme. Prior to the segmentation, a normalization scheme as prepro cessing step is used to reduce illumination gradients, which is also able t o treat illumination, edges like shades. (C) 2001 Elsevier Science B.V. All rights reserved.