Estimators for Orientation and Anisotropy in Digitized Images

摘要: This paper describes a technique for characterization and segmentation of anisotropic patterns that exhibit single local orientation. Using Gaussian derivatives we construct gradient-square tensor at selected scale. Smoothing this allows us to combine information in neighborhood without canceling vectors pointing opposite directions. Whereas would cancel, their tensors reinforce. Consequently, the characterizes orientation rather than direction. Usually is least few times larger scale parameter gradient operators. The eigenvalues yield measure anisotropy whereas eigenvectors indicate In addition these measures can detect anomalies textured patterns. 1. Introduction Information from subsurface structures may help geologists search hydrocarbons (oil gas). seismic measurements which are performed earth’s surface important be extracted borehole. done either by downhole imaging borehole wall or analyzing removed material “the core”. Core requires careful drilling with hollow drillbit. cores transported further analysis. Apart physical interested spatial organization acquired rock formations. We show quantitative image cylindrical cut longitudinally (slabbed) digitization flat yields 2D slabbed core image. Quantitative about layer structure geologist improve interpretation. approach followed guided simple model subsurface. These layers described number parameters all vary as function depth. Some have direct geometric meaning (dip azimuth) others much more difficult express quantitatively unique way. will focus on applied images.