Contour 3D

Overview

This library performs 3D contouring by extracting iso-surfaces from three-dimensional arrays and outputting the result as a portable pixmap (ppm) image.

The objective is to visualise three-dimensional flow fields simulated on massively parallelised CFD solvers (available at here) without saving the full three-dimensional arrays. By embedding this library into the existing flow solver, one can easily create movies without worrying about storage limitations.

This library does not have any additional dependencies on X, Qt, VTK, OpenGL, etc., making the initial cost minimal.

Caveat

The motivation of this project is to visualise the (flow) fields quickly and intuitively without the support of graphical libraries. For beautiful and comprehensive renderings, use other software options that offer more features.

Dependency

C compiler
GNU Make
MPI
Simple Decomp Library (included as a sub-module)

Quick start

Prepare workplace

mkdir -p /path/to/your/directory
cd       /path/to/your/directory

Get source

For example:

git clone --recurse-submodules https://github.com/NaokiHori/Contour3D
cd Contour3D

Do not forget to fetch the sub-module as well.

Build
```
make clean
make all
```
Execute
```
mpirun -n 2 --oversubscribe ./a.out
```
This may take a few seconds. Change the number of processes depending on your machine spec.
Check output

Find output.ppm, which is the result of the 3D contouring:

Detail

See src/main.c to investigate how the contours, the camera, the light, and the screen are configured.

Basically this library does the following things:

Array extension

Edges of the given three-dimensional array are communicated among all processes to avoid gaps.
Tessellation

From the extended three-dimensional array, each process extracts triangular elements and their surface normals using the marching-tetrahedra algorithm.
Smoothing

Vertex normals are computed by averaging the surface normals of the neighbouring triangles to obtain a smoother result.
Rendering

Decide the color of each element based on the direction of the light and the local normal vector interpolated on each barycentric coordinate.
Reduction

Among all processes, find the nearest triangular element to the screen and output the result to an image.

The steps 1-4 are repeated if multiple arrays and/or thresholds are given.

See src/contour3d/main.c to check the overall procedures.

Reference

Acknowledgement

I would like to thank Prof. Roberto Verzicco for a stimulating lecture in the JMBC course Multiphase Flow and Phase Transitions.