Disparity to Z Node

Disparity to Z applies the geometric relationship between disparity and depth: the larger the disparity, the closer the object. The exact conversion requires the parameters of the stereo rig: the distance between the eyes (interocular), the convergence point and the focal length.

Output: a scalar Z-depth you can use like any other Z (Depth Blur, Fog, Volume Mask, etc.). It gives you 3D-aware effects on live action stereo footage without producing the Z by other means.

• Effects Library: Tools > Stereo > Disparity to Z
• Quick add: right-click in the Node Editor → Add Tool > Stereo > Disparity to Z

In epipolar geometry (parallel cameras), the relationship is:

Z = (B × f) / d

Z = depth (camera → object distance), B = baseline (interocular, the distance between the cameras), f = focal length, d = disparity in pixels.

Consequence

At a constant disparity, raising B or f increases the perceived depth. With B and f fixed, high disparities correspond to low Z values (close objects).

Convergence

The distance at which the optical axes cross. Defines the "screen plane": disparity = 0 on that plane, negative in front of it, positive behind it.

• Disparity (orange) - an image containing the disparity map (output of a Disparity node).
• Effect Mask (blue) - optional.

Stereo → Z-depth workflow:
LeftCam ─┐
         ├── Disparity → Disparity to Z (real camera params)
RightCam ┘                ↓
                        Z-depth map → Depth Blur, Fog, etc.

Stereo depth of field in post:
LeftCam ─┐
         ├── Disparity → Disparity to Z
RightCam ┘                ↓
        ┌──────────── Depth Blur (Focus distance, Aperture)
        └── Beauty L → DOF applied based on Z
                       ↓ MediaOut

Volume Fog on stereo footage:
LeftCam ─┐
         ├── Disparity → Disparity to Z → World (with camera params)
RightCam ┘                                  ↓
                                        Volume Fog
                                          ↓ MediaOut

Interocular (default: 6.5 cm)

Distance between the two cameras (= baseline B). 6.5 cm is the average distance between human eyes. For specific rigs, measure it physically.

Convergence (default: 0)

Distance of the convergence plane (where disparity = 0). Defines the screen plane. Get this value from the rig metadata or the shooting notes.

Focal Length (default: 35 mm)

The optical focal length in mm. Critical for precision: a mismatch distorts every Z value. Read it from the EXIF data or the camera notes.

Sensor Size X / Y (defaults: 35 mm × 24 mm)

Sensor dimensions. Together with the focal length, they determine the effective angle of view. Full frame = 35×24, Super 35 = 24.89×18.66, etc.

Output Mode (default: Distance)

Distance = Euclidean distance, Linear Depth = linear Z component. Choose based on the downstream nodes (Volume Fog prefers Distance, Depth Blur prefers Linear depending on the convention).

• DOF in post-production on stereo - fully in-focus stereo footage, add controllable bokeh in post with a Z map derived from the disparity.
• Volume Fog on live action stereo - a 3D-aware atmosphere on real footage shot on a stereo rig.
• 3D compositing in a stereo plate - a Z map to position and correctly occlude added 3D elements.
• Color grading by depth - different grading depending on distance (warm close-ups, cool distance) using the derived Z.
• Generating mattes by depth - a matte of the background or the foreground by thresholding the Z map.
• Clean plate / background removal - combine the Z map + roto to isolate the subjects from the distant background.

• Exact camera parameters - Interocular, Convergence and Focal Length must match the real shooting values. Otherwise, the Z is distorted. Get them from the rig's spec sheet.
• Check the visualization - the visualized Z output should show a clean gradient from near (bright) to far (dark), or the reverse depending on the convention.
• A clean disparity is essential - artifacts in the disparity map carry over into the Z. Clean up the disparity (suitable Smoothness) before this conversion.
• 32-bit float is mandatory - Z values go far beyond 0-1. In 8/16-bit you lose all precision.
• Convergence = the screen plane - pixels with a disparity of 0 in the source = at the convergence plane in the output. So their Z = the Convergence value of the node.
• Sensor Size for anamorphic - be careful with anamorphic shoots: adapt the Sensor Size to the non-standard ratio, or desqueeze before Disparity.
• For workflows with no camera information, make rough visual estimates, then adjust by comparing against an object of known depth in the viewer.