Rendering is computing the image of a scene: given geometry, materials, lights and a camera, decide the colour of every pixel. Real-time renderers trade accuracy for speed to hit a frame every few milliseconds; offline renderers simulate light more faithfully and may take minutes or hours per image.
Key concepts
- Rasterisation: project triangles onto the screen and shade the pixels they cover; fast, and the basis of real-time graphics.
- Ray tracing: follow rays from the camera through the scene to find what they hit and how light reaches it.
- Shading models: diffuse and specular reflection, and physically based materials that conserve energy.
- Global illumination: light bouncing between surfaces, which gives soft shadows and colour bleeding.
- Sampling and noise: Monte Carlo methods estimate light with random samples; more samples mean less noise.
Vocabulary
- BRDF: a function giving how much light a surface reflects from one direction to another.
- Path tracing: ray tracing that follows random light paths to solve the rendering equation.
- Shadow map: a depth image rendered from a light, used to test what is in shadow.
- Denoising: filtering a noisy, low-sample image into a clean one.
What to study, in order
- Write a small ray tracer for spheres and planes that writes an image file.
- Add diffuse and specular shading, shadows and reflections.
- Turn it into a path tracer; watch noise fall as samples grow.
- Implement a triangle rasteriser with a depth buffer and compare the results and the cost.
- Read about physically based materials and implement a basic metal and dielectric.
- Add acceleration structures such as a bounding volume hierarchy and measure the speedup.