An orientation and resolution independent texture classifier in segmentation of images of unknown rotation and scale

Describes an orientation and resolution insensitive texture classifier for demarcating regions of common texture within a rotated and/or scaled image. The algorithm involves a mask `tuning´ process performed over equal size pure texture samples in the multi-orientation and multi-scale data set via a two-dimensional linked-list. The development of a guided random search approach for parameter optimization is detailed and the segmentation results of collages of Brodatz textures are provided