depthToSurface

DepthToSurface refers to a technique used in computer graphics to convert a value from a depth buffer into a 3D point on the corresponding surface along the camera ray. It enables reconstruction of a surface or scene geometry from depth data, which can then be used for shading, lighting, or post-processing effects.

The core idea is to take the per-pixel depth and the pixel’s screen coordinates and recover the 3D position that produced that depth. In practice, this usually involves forming a 4D clip-space vector from the pixel’s normalized device coordinates and depth, then multiplying by the inverse projection matrix to obtain a point in camera space, and finally applying the inverse view (and optionally the inverse world) transform to yield a position in world space. Depth values are often non-linear with perspective projection, so some implementations linearize depth before certain computations.

Implementations require information about the camera, including its projection parameters or matrices, the viewport size, and the coordinate of the pixel being reconstructed. Some pipelines perform reconstruction directly in view space, while others convert to world space. Different graphics APIs may store depth in different ranges (for example, DirectX versus OpenGL), and depthToSurface routines account for these conventions.

Applications of depthToSurface include screen-space effects such as ambient occlusion, reflections, and depth-based shading, as well as depth-image-based rendering and point-cloud-like reconstructions from a depth map. Limitations include depth buffer precision, non-linear depth distribution, and potential artifacts near depth discontinuities or at object silhouettes. Proper handling of projection parameters and depth ranges is essential for accurate reconstruction.