A set of multiresolution texture features suitable for unsupervised image segmentation

摘要: We propose a set of multiresolution features for texture description. Image smoothing at multiple scales using the fast B-spline transform is performed and number features, such as local area, normal vector dispersion gradient orientation, are computed from each scale. A simple disparity function applied to assess discriminative power these with comparison other methods. Being effective even small observation windows, proposed suitable high-resolution segmentation.