Automatic Rock Classification Based on Texture Analysis in Electrical Borehole Wall Images

Electrical borehole wall images are widely used for borehole inspection and reservoir characterization. So far, this data is mostly qualitatively used to investigate structure and lithology mapping. We present a method for image characterization which is based on the application of texture analysis in order to transform image data into quantitative log curves. We derive so-called Haralick texture features from borehole wall images. Based on a supervised classification technique, we train texture features within assigned rock classes, determine their classifiers, and apply classification function on the entire data set. This enable automatic rock determination based on quantitative image data.