This was my term project in the Introduction to Digital Rock Petrophysics course. For this project I used the Savonnières carbonate dataset from the Digital Rocks Portal: https://www.digitalrocksportal.org/projects/72 The original image stack is shown below:
The slices in the image stack were then aligned using the Enhanced Correlation Coefficient algorithm (aligning_code_4). Prior attempts using phase correlation (aligning_code_1,2,3) yielded unsatisfactory results regardless of format. Parallel processing was used to improve the speed of the process. The image was also scaled down by a factor of half. Resulting stack is shown below:
The scaled and aligned image stack was then cropped using Fiji-ImageJ:
Otsu method within Fiji-ImageJ was then used to threshold the image. Pores were given voxel values of 0 and minerals 255. The final scaled, aligned, and thresholded image stack is shown below. The 2D image stack and white 3D representation were generated using Fiji-ImageJ. The coloured 3D model was generated using Paraview.
Monte Carlo sampling was then used to generate random cuboid 3D samples from the image stack. This was to enable the calculation of certain petrophysical parameters that required cuboid 3D images.
I used a random sample generator with a fixed seed. For each seed I tried I manually went through the samples and determined which of the samples were "high-value" i.e. did not touch the edge of the original image stack and did not include the outside of the image stack.
The code used for their generation is "random_sample_gen2" The parameters I used to generate the random samples is in the code. The number of high-value samples versus seed number is outlined in "good samples.txt" 25 samples were generated and the 11 samples chosen are shown in the table below:
Paraview was then used to view the 25 samples generated in 3D as shown below:
The high-value samples selected are shown in red:
These were then used to calculate the petrophysical properties of the Savonnières carbonate.