Ray to Light Field Coordinates

mediumPROLight FieldsLight FieldCV: Image-Based Rendering

Convert a 3D ray to light field (s, t, u, v) coordinates.

A light field parameterizes all rays in a scene using the two-plane representation:

$(s, t)$ : intersection with the first plane (camera plane)
$(u, v)$ : intersection with the second plane (focal plane)

Given a ray starting at origin $(x_0, y_0, z_0)$ on the first plane (at $z=0$ ) with direction $(d_x, d_y, d_z)$ , find where it intersects the second plane at $z = z_{plane}$ :

$t_{param} = \frac{z_{plane}}{d_z}$ $u = x_0 + t_{param} \cdot d_x$ $v = y_0 + t_{param} \cdot d_y$

Example

Input

ray_to_lightfield([0, 0, 0], [0, 0, 1], 1)

Output

(0, 0, 0, 0)

Reasoning

Inputs: ray_origin is the point $(x_0, y_0, z_0) = (0, 0, 0)$ where the ray leaves the camera plane $z = 0$ . ray_direction is $(d_x, d_y, d_z) = (0, 0, 1)$ , pointing straight down the $z$ axis. plane_distance puts the focal plane at $z_{plane} = 1$ .
Camera plane: $(s, t)$ is just where the ray already sits on the first plane, so $(s, t) = (x_0, y_0) = (0, 0)$ . The $z_0 = 0$ part carries no information.
Reaching the focal plane: the ray climbs $d_z = 1$ in $z$ per unit step, so it needs $t_{param} = \frac{1}{1} = 1$ steps to reach $z = 1$ .
Focal plane: $u = 0 + 1 \cdot 0 = 0$ and $v = 0 + 1 \cdot 0 = 0$ , rounded to 4 decimal places.
Output: (0, 0, 0.0, 0.0). The ray hits both planes at their centres.
Notice that $(u - s, v - t)$ encodes the ray's angle: a tilted direction like $(1, 1, 1)$ with the same origin lands at $(u, v) = (1, 1)$ instead.

Constraints

ray_origin: [x, y, z] point on first plane (z=0)
ray_direction: [dx, dy, dz] direction vector
plane_distance: z-coordinate of second plane
Return (s, t, u, v) coordinates, rounded to 4 decimal places

Learn this in

Image-Based RenderingCV Study Plan

More from CV: Image-Based Rendering

Differentiable Renderer Forwardhard Video Loop Findinghard Light Field Refocusinghard Depth-Based View Synthesishard Front-to-Back Splat Accumulationmedium

Concept primerPREVIEW

Ray to Light Field Coordinates: Background & Strategy

Background Knowledge

Light Field Representation and the Two-Plane Model

A light field is a complete description of all light rays in a scene, capturing both the intensity and direction of light traveling through 3D space. The two-plane parameterization (also called the 4D light field or Lumigraph representation) is a practical way to encode this information using just four coordinates. The first plane (typically the camera plane at $z=0$ ) records where a ray originates or passes through, while the second plane (the focal plane at some distance $z_{plane}$ ) records where that same ray intersects further along its path. This representation is fundamental to image-based rendering because it allows you to synthesize novel viewpoints by resampling the light field at different $(s,t)$ positions.

Ray Parameterization in 3D Space

In computer graphics and computational imaging, rays are typically represented parametrically: a ray starts at a point $(x_0, y_0, z_0)$ and travels in direction $(d_x, d_y, d_z)$ . Any point along the ray can be expressed as: $\mathbf{p}(t) = (x_0, y_0, z_0) + t \cdot (d_x, d_y, d_z)$

where $t$ is a scalar parameter (often called the "time" or "distance" along the ray). To find where a ray intersects a plane perpendicular to the $z$ -axis at depth $z_{plane}$ , you solve for the parameter value where the $z$ -coordinate equals $z_{plane}$ , then use that parameter to compute the $x$ and $y$ coordinates at the intersection point.

Coordinate Mapping and Perspective Projection

The conversion from 3D ray space to light field coordinates is essentially a form of perspective projection. The $(s,t)$ coordinates represent the ray's position on the first plane, while $(u,v)$ represent its position on the second plane. This mapping encodes both the spatial location and the direction of the ray in a way that's efficient for rendering and analysis. Understanding this relationship is crucial for tasks like view synthesis, depth estimation, and light field reconstruction.

Algorithm/Approach

The solution follows a straightforward ray-plane intersection pattern:

Parameterize the ray using the given origin and direction
Solve for the intersection parameter by setting the ray's $z$ -coordinate equal to $z_{plane}$
Compute the intersection point by evaluating the ray equation at that parameter value
Extract the light field coordinates by reading off the $x$ and $y$ components of the intersection point

PRO

Continue the full explanation

You're reading the free preview. Unlock the complete walkthrough, the code editor, test runner and reference solution with Premium.

PRO

Editor locked

The code editor is locked for Pro problems. It is only available for free problems. Please upgrade to gain access to the code editor for all problems.

solution.py

Python 3.13

Test results