# Projector edge blending and overlap

Learn how projector overlap, geometric alignment, blend curves and light response work together, with practical seam diagnosis for immersive screens and domes.

Canonical: https://docs.theusualnext.com/en/tholusos/projection/edge-blending/
Updated: 2026-10-22
Publisher: the usual neXt

**Edge blending** progressively attenuates images where multiple projectors illuminate the same surface. Its purpose is to make their combined light uniform. A correct seam also needs geometric alignment, consistent light response and separate black-level adjustment.

These principles apply to panoramic walls, domes and immersive rooms. In TholusOS, blending is part of [calibration](/en/tholusos/projection/calibration/) and follows geometric map generation.

## Separate geometry, blending and black level

Warping determines where a pixel appears. Blending determines how much light each projector contributes. Black-level compensation addresses residual light in dark scenes. Treating these as separate problems helps avoid fixing one test image while making another worse.

| Problem | Relevant control | Useful pattern |
| --- | --- | --- |
| A line doubles in the overlap | Geometry and alignment | Fine-line grid |
| The seam is bright or dark in midtones | Blending, light response and uniformity | Uniform grey at several levels |
| The seam changes colour | Picture presets and RGB response | Neutral grey and colour fields |
| The seam appears in near-black scenes | Black level and residual light | Black and low-grey patterns |

NVIDIA likewise describes warp and blend as complementary image-geometry and intensity corrections for multiprojector displays. [NVIDIA, Warp and Blend](https://developer.nvidia.com/warp-and-blend).

## Design physical overlap first

Blending requires a region that both projectors actually illuminate. Software distributes light within that region; it cannot add coverage beyond the beam. Assess overlap, focus and useful resolution before finalising projector positions.

For a flat wall with two equally scaled images, a simple arithmetic example is:

```text
useful width = width A + width B − overlap
             = 1920 + 1920 − 240
             = 3600 pixels
```

This ideal example assumes equally sized pixels on the screen, no cropping and 240 pixels of overlap. Dome curvature introduces varying angles and pixel density. Adding the projectors' nominal resolutions does not describe the actual visible detail.

There is no universal overlap percentage. A narrow region leaves less room for alignment and transition. A wide region spends more pixels on the same area and combines more residual light in black scenes. Evaluate the optical design and difficult content before choosing the compromise.

Inspect intersections involving three or more projectors too. A system that looks correct at each two-projector seam can still have a bright spot where several beams meet.

## Combine light correctly

Blend weights must divide the required light contribution. In an ideal model with two identical, linear projectors, complementary weights satisfy `wA + wB = 1`. Real projector response matters because halving a digital signal does not necessarily halve emitted light.

NVIDIA's gamma discussion explains the distinction between encoded values and linear light: light calculations need a linear representation, while image encoding and display response may be nonlinear. [NVIDIA GPU Gems, The Importance of Being Linear](https://developer.nvidia.com/gpugems/gpugems3/part-iv-image-effects/chapter-24-importance-being-linear).

As a simplified mathematical example, if `L = V^2.2`, halving the signal from two white images produces:

```text
2 × 0.5^2.2 ≈ 0.435
```

That is about 44% of the reference light instead of the intended 100%. This explains a model; it is neither a projector measurement nor a recommended gamma setting. Real response may change with picture mode, signal level, colour channel and light-source power.

Residual black adds another term that multiplication by a digital blend weight cannot remove. Calibration therefore also needs to address [black and low-grey levels](/en/tholusos/projection/black-level/).

## Adjust blending in TholusOS

The light-and-seam stage uses the overlaps derived from the warp maps. It offers light measurement or explicit confirmation of the base blend. Complete alignment and the visible-area boundary first: the fade width can only work within the geometry already available.

1. Inspect a grid and correct doubled images or discontinuities before adjusting brightness.
2. Check that the projectors are stable and that picture modes, light modes and automatic settings agree with the capture conditions.
3. Generate blend maps and evaluate the proposed light optimisation.
4. Explicitly apply a usable measurement result. If you choose the base blend, inspect the result without treating it as a successfully measured light correction.
5. Test light, medium and dark grey fields, then proceed to black-level adjustment and final video verification.

The blend curve changes the distribution of the transition. A seam that stays too bright or dark at several levels can require light-response correction rather than repeated curve changes. Incomplete warp coverage requires a geometry fix.

## Diagnose a visible seam

The quickest useful diagnosis compares the same location under different patterns and records when the defect appears. Keep the camera, room lighting and projector settings fixed throughout the comparison.

| Observation | Hypothesis to check |
| --- | --- |
| The defect remains with one projector | Focus, optics, screen or individual-projector uniformity |
| Lines align but grey changes across the seam | Light response or blending |
| Mid-grey matches but black is brighter | Added residual light |
| The seam pulses as the scene changes | Active dimming or dynamic contrast |
| A colour cast remains in individual images | Picture preset, source or colour-channel differences |
| A sharply defined black strip | Missing coverage, visible-area mask, slot or map mismatch |

Finish with slow movement, dark scenes and representative programme content. An automatically exposed photograph or a preview on the Mac display can hide differences visible in the room. Record test conditions and results so later changes can be recognised.