3-D texture characterized by accessibility measurements, based on the grey weighted distance transform
Karel C. Strasters, Arnold W. M. Smeulders, Hans T M van der Voort · Bioimaging · 1994
A method of extracting texture information from a 3-D grey level distribution is proposed, based on the concept of accessibility. We apply this method to cell nuclei in confocally scanned images. Inaccessibility of a point is defined as the minimal sum of optical density that is encountered when traversing from outside the object of interest to this point. The inaccessibility of all object voxels can be determined by applying a Grey weighted Distance Transform (GDT). From the inaccessibility patterns three texture features are extracted: object inaccessibility, object inaccessibility compared to a homogeneous distribution, and object homogeneity. Normalization of these features enables objective inter object comparison of texture. Simulations on artificial textures in varying object shapes show the shape independence and discriminating power of the proposed texture features. The processing steps that are necessary for confocally scanned data are discussed and an application of analysing a cell nucleus is shown.