Image analysis applied to quartz grain microtextural provenance studies

Abstract

Every quartz grain surface is sculptured during time by many different processes. During the life of a grain, sediment transport, physical and chemical processes imprint specific microtextural features on their surface. State-of-the-art methods for analysis of the surface of quartz particles in Scanning Electron Microscope images are performed non-automatically by experts and essentially are dependent on operator proficiency. Here, we propose digital image processing-based method for automated microtextural analysis of grain's surface with the goal to classify the grain into specific classes corresponding to its genesis (last provenance cycle). The method is based on original classification scheme which extracts the textural features from the digital images of single grains in training database and creates Texton Dictionary for establishing the prototype of specific texture set characterizing given genesis class. The created Texton Dictionary then can be used for classification of a grain with unknown genesis with very good performances when tested on databases of grains with known origins.