Post-production and mastering
Before it reaches the viewer, the stereo pair shot in the field is corrected in post-production, its depth is adjusted to the target screen, and a separate final copy (master) is prepared for each target screen. On this page, "shot" means a single camera shot. Field calculations often do not know the target screen, or assume a single screen; mastering closes this gap.
Geometric correction
Only horizontal differences between the two images carry depth information. Every other difference is corrected (rectification):
- vertical alignment and rotation differences between the camera axes,
- scale differences caused by differences in focal length or zoom,
- the keystone and depth plane curvature of converged shooting,
- distortion differences between the two lenses.
Color and brightness matching
Color, brightness and contrast must be the same for both eyes. Differences come from the lenses, the sensors and, especially in a mirror rig, from the mirror affecting transmitted and reflected light differently. Areas such as bright reflections and the sky are checked separately.
Depth grading
Where the shot sits relative to the screen is set by horizontal image translation (HIT). HIT shifts all parallax by the same amount and does not change the depth range:
where: d parallax (as a fraction of image width), H HIT amount, d_min and d_max the parallax of the nearest and farthest point of the shot
HIT is equivalent to changing the convergence in the field:
where: C convergence distance in the field, C' equivalent convergence after HIT
Target screen masters
Each screen has its own comfort window: the narrowest of the screen-type budget, the divergence limit and the 1° angular limit (see Comfort zones). If a shot fits a screen with HIT:
where: pos and neg the back and front limits of that screen's comfort window
If the HIT ranges of several screens overlap, a single master covers them all; if not, the screens are grouped into the fewest masters. If the range does not fit in the window, depth must be compressed:
Depth compression cannot be done by shifting the images alone; it requires intermediate view generation or a reshoot. So for a screen that does not fit, the measure is the equivalent interaxial that should be used in the field.
Roundness also changes with the target screen, because it depends on the viewing distance:
where: V_k the viewing distance of that screen
Example
35 mm lens, Super 35 sensor, interaxial 60 mm, convergence 3 m, scene between 2 m and infinity. Parallax −1.41% … +2.81%, range 4.22%:
| Target screen | Comfort window | Result |
|---|---|---|
| Cinema, 10 m screen, 15 m | −2.00% … +0.65% | Does not fit; ×0.63, equivalent to b 37.7 mm in the field |
| TV 55", 2.5 m | −1.00% … +2.00% | Does not fit; ×0.71, equivalent to b 42.7 mm in the field |
Editing continuity
At cuts, the depth of the main subject should not jump suddenly. In consecutive shots the parallax of the main subject is kept close; if needed, HIT is changed gradually before and after the cut. See Vergence–accommodation conflict.
Window violations
For objects that come in front of the screen and touch the edge of the frame, a floating window is applied. When HIT changes, the violation and the required mask width change too; so the floating window is set after depth grading.
Erkan Cerit Method
The method set: formula used in the field + S3D Mastering = Final Mastering. Its field part is the Cerit Triangle. In the Cerit Triangle, the part that extends beyond the comfort window stays outside the in-focus range, that is, blurred; so even if the whole scene does not fit on a screen, the shot counts as balanced if the in-focus range fits. The screen-by-screen results of the same shot are calculated on the Mastering page.