Retrieving scale from quasi-stationary images

摘要: We have developed a novel method to derive scale information from quasi-stationary images, which relies on rotation-guided multi-scale analysis of features derived Gray-Level Co-occurrence Matrices (GLCM). Unlike other methods for texture characterization, our does not require rotation-invariant textural features, but instead uses orientation the image constrain algorithm. Our computes GLCM ''stencil'' that follows local field. It compares obtained sliding window scans whole with those present user-selected reference pattern. The then calculates similarity measure between and By applying different affine transforms stencil used sampling pattern, we are able regions dilated versions hence perform multi-resolution image. For given region an image, is find most likely scale. Therefore it can estimate stationarity in terms scale, has important applications multipoint geostatistics (MPGS). tested Brodatz textures database. multi-scale, algorithm derives images. extends variable size, oriented, image-sampling ''stencils'', measures patterns full achieves successful MPGS.