Skew Divergence-Based Fuzzy Segmentation of Rock Samples

Digital image segmentation is a process in which one assigns distinct labels to different objects in a digital image. The MOFS (Multi Object Fuzzy Segmentation) algorithm has been successfully applied to the segmentation of images from several modalities. However, the traditional MOFS algorithm fails when applied to images whose composing objects are characterized by textures whose patterns cannot be successfully described by simple statistics computed over a very restricted area. Here, we present an extension of the MOFS algorithm that achieves the segmentation of textures by employing adaptive affinity functions that use the Skew Divergence as a measure of distance between two distributions. These affinity functions are called adaptive because their associated area (neighborhood) changes according to the characteristics of the texture being processed. We performed experiments on mosaic images composed by combining rock sample images which show the effectiveness of the adaptive skew divergence based fuzzy affinity functions.