This project provided an appreciation of how 3D graphics APIs such as DirectX and OpenGL work under the covers. To do this, a software renderer (written in C++) was built up from scratch. The following concepts were covered:
- Geometric Transformations — How to get from model co-ordinates to screen co-ordinates a.k.a. the viewing pipeline.
- Vector maths — Calculating vectors from points, vector addition, vector length, dot product and cross product.
- Lighting — Ambient, directional, point and spot lighting.
- Reflection — Diffuse and specular (Phong/Blinn-Phong reflection models).
- Rasterisation — Gouraud shading.
- Cameras & Viewpoints — The view frustum, polygon culling and clipping.
The evolution of Marvin image below demonstrates the progression of the renderer. It starts with a simple wireframe, progressing to flat shaded polygons (using GDI+), and ending with a custom Gouraud shader to smooth out the image. Lighting effects are then applied to modulate the overall appearance, and finally the model is textured.
The full evolution can be seen in the video along with an explanation of what is being demonstrated at each stage.
This was probably the most challenging project I have worked on to date, but ultimately it proved very rewarding as some really nice effects can be achieved. I really enjoyed the modular nature of how the project was built up, with a study of a particular algorithm, followed by its implementation. This allowed the application to be built up gradually, preventing it from becoming too overwhelming.