The promise of Image-Based Rock Physics (IBRP), often referred to as Digital Rock Physics, has been around since the 1950s when simple pore networks of nodes and connectors were analyzed with electrical resistor networks. The intervening 60+ years has produced significant developments in imaging and computational power to where IBRP is used to predict multiphase fluid flow in complex pore systems. 

IBRP has two roles in today’s Oil & Gas industry:

  1. As a Predictive tool in rocks that are difficult to measure in the lab.
  2. As an Evaluation tool for complex SCAL measurements where a series of numerical experiments that vary input parameters can be executed faster than comparable lab tests.



Comparison of Image Segmentation from “Uncertainty Quantification in Image Segmentation for Image-Based Rock Physics in a Shaly-Sandstone”  For Sample B, 4-micron resolution, 680x680x900 volume, slice 450. 

DigiM’s I2S software platform returns control of the IBRP process to the client by providing a state-of-the-art image segmentation tool based on Machine-Learning principles coupled with a series of simulation modules that capture a range of single and multi-phase flow properties.

IBRP includes contributions from three separate disciplines:

  1. Images of the pore geometry at the required resolution to capture critical features,
  2. Image processing to segment the pore geometry from the rock matrix,
  3. Simulation of petrophysics properties, including single- and multi-phase transport phenomena.

In the DigiM I2S platform, the acquisition of images is returned to the client. DigiM can provide guidance on the best methods to use, but recognizes that the client often has more expertise on how to image their own samples. 

The ML-based image processing tool does reduce the amount of pre-processing of the image that is required before segmentation steps. Traditional image pre-processing of smoothing, filtering, edge enhancements are less important now as ML-based segmentation is much more robust and can overlook image defects. The ease in training the ML-segmentation tool allows the client real-time information on the quality of their segmentation setup. Multiple grain and pore types can be identified reliably with the result of capturing more what the users’ eyes observe in the images.


The I2S platform includes a suite of simulation modules that cover a range of petrophysical properties that include permeability, electrical and thermal conductivity, and basic characterization of the pore space such as pore size, capillary pressure, and tortuosity. 

The modules use standard algorithms and numerical solvers, but I2S returns control of the simulation to the clients. Input parameters are client-selected, which encourages numerical experimentation on the importance of individual contributors to the simulation results.