Rasterization converts projected geometric primitives into samples on a raster image.

For a projected triangle, the rasterizer first determines which pixel sample locations or pixel areas are covered by the triangle. Barycentric or edge-function tests provide an efficient inside-triangle criterion.

For every covered location, the renderer produces a fragment carrying interpolated attributes such as depth, color, normal and texture coordinates. A fragment is a candidate contribution to a pixel; later visibility and blending tests determine whether and how it changes the image.

Rasterization is a forward rendering method: geometry is projected toward the image and asks which pixels it covers. This contrasts with ray tracing, where image samples ask which scene geometry is visible along a ray.

Shared-edge rules must be consistent so adjacent triangles neither leave cracks nor shade the same boundary unpredictably.

Rasterization does not itself decide lighting or visibility; it creates the per-pixel candidates on which those later operations act.