Reservoir Rocks Image Binarization: a Comparative Study Between Neural Network Method and Otsu Variance
Eneida Arendt Rego, André Duarte Bueno · 2015
The determination of physical properties of reservoir rocks is one of the most important tasks in reservoir engineering. These properties can be determined through microscopic image processing of porous media. The big advantage is the low cost and greater speed in obtaining results. Having this background as motivation and inspired by the success of artificial intelligence systems, this work analyzes the results of binarization images of reservoir rocks using histogram-based methods described in the literature and the development of a method of image binarization based on neural networks, which is a specific method for images of reservoir rocks. As a result, binarized images are displayed and their characterization for different binarization methods and the efficiency of the methods evaluated. Binarization using neural networks showed an improvement of 6.8% to 71.0% compared to Otsu Variance.