Rasterization and the Graphics Pipeline

Rasterization converts geometric primitives such as triangles into the pixels they cover on screen, and forms the central step of the graphics pipeline that transforms 3D scenes into 2D images.

Definition

Rasterization is the process of determining which pixels a projected geometric primitive covers and interpolating per-vertex quantities such as depth, color, and texture coordinates across those pixels.

Scope

This topic covers the stages of the standard rendering pipeline - vertex transformation, clipping, projection, primitive assembly, scan conversion, and fragment processing - along with depth buffering for visibility, perspective-correct interpolation of vertex attributes, and antialiasing of the resulting samples.

Core questions

• How are 3D coordinates transformed into 2D screen positions?
• Which pixels does a projected triangle cover?
• How is visibility resolved when primitives overlap?
• How are aliasing artifacts along edges reduced?

Key concepts

• Vertex and fragment stages
• Clipping and projection
• Scan conversion
• Z-buffer depth test
• Perspective-correct interpolation
• Antialiasing

Key theories

The transform-and-rasterize pipeline

Geometry passes through a fixed sequence of coordinate transformations from object space to screen space, after which primitives are scan-converted to fragments, providing a structure that maps efficiently onto parallel hardware.

Depth buffering for visibility

The z-buffer stores the nearest depth seen so far at each pixel and discards fragments that lie behind it, resolving hidden-surface removal incrementally without sorting geometry.

Clinical relevance

The rasterization pipeline is the foundation of virtually all real-time graphics, driving video games, user-interface compositing, CAD viewers, and the interactive 3D found in browsers and mobile devices.

History

The z-buffer and scan-conversion methods of the early 1970s were codified into the fixed-function pipeline of early graphics hardware, then generalized into the programmable shader pipelines of modern GPUs.

Key figures

• Edwin Catmull
• Bui Tuong Phong

Seminal works

• catmull1974
• hughes2013

Frequently asked questions

Why is rasterization so fast?

Each triangle is processed independently and each pixel is filled by simple incremental arithmetic, which maps directly onto the massively parallel architecture of graphics hardware.

What is the z-buffer for?

It records how far away the closest surface drawn so far is at each pixel, so that nearer surfaces overwrite farther ones and hidden parts of the scene are automatically removed.