Red, green and blue filters give you the color a nebula really has. A hydrogen-alpha filter gives you far more of its structure. It cuts through light pollution and moonlight and records the faint hydrogen that broadband red barely registers.

HaRGB gives you both: a natural-color image whose nebula is as deep as the Hα data allows. It is also easy to do badly.

The idea

Hα at 656.3nm is deep red, well inside the passband of a red filter. So hydrogen light is already in your red channel, but faintly, diluted by everything else red lets through, including the sky glow. The Hα stack holds the same light, far cleaner.

HaRGB puts that cleaner hydrogen signal into red, where it belongs, and leaves green and blue alone. The result stays true color: stars keep their colors, the sky stays neutral, and the nebula is the red it really is, with more of it showing.

Two ways it goes wrong

Averaging Hα with red. This is the obvious move, and it halves both. The nebula comes out no stronger than before, and every star in the red channel dims, because Hα stars are faint and small. Stars drift toward cyan.

Adding Hα to red. Better nebula, but the sky doubles too: noise and sky glow from both frames pile up, and the whole background goes red-brown. You then spend the rest of the edit removing a cast you added yourself.

What works: the lighten blend

Red takes Hα only where Hα is brighter than red, which is where there is hydrogen emission the broadband filter recorded only faintly. Everywhere else, red stays red.

Two refinements matter in practice:

Compare above the noise. A plain “take the brighter” of two noisy images is biased upward by about half a noise-sigma everywhere, which paints the whole sky faintly red. Taking only the part of Hα’s lead that exceeds the noise leaves the sky where the color balance put it.
Do not take all of it. On linear data the hydrogen lead can be large. Blended in full, the nebula saturates to flat, clipped red, and the brightest ridge can lose its color entirely. In Akastroid the blend adds about a third of Hα’s lead. That figure was chosen by rendering several strengths side by side against the plain stack; at a third the nebula keeps the stack’s red-magenta and nothing clips.

Balance first, blend second

The broadband color has to be balanced on the stars before the hydrogen goes in. In a frame full of nebula, red’s brightest pixels are nebula while green’s and blue’s are stars; scale each channel to its own brightest pixels and red ends up scaled down against the others, so faint stars go green. Balance R, G and B on the stars, give Hα the same scaling red received, and only then blend.

The combining itself has to happen on linear stacks, before any stretch. Stretch each channel on its own first and a star’s color starts to depend on how bright it is, because each channel went through a different curve. That is covered in why stars go yellow or green in RGB composites.

When it is worth doing

Emission nebulae (the North America, the Heart, the Rosette, M42) gain the most. There is a worked example in processing the North America Nebula.
Galaxies gain a little: their HII regions show as pink knots along the arms.
Star clusters and reflection nebulae gain nothing, because there is no hydrogen emission to add.

Since version 1.3.2, Akastroid does this when a filter-wheel folder holds Hα alongside R, G and B: it offers Natural color + Hα as the default picture, with the Hubble palette and HOO beside it. The mono and narrowband page covers the rest of the filter-wheel workflow.

The short version

HaRGB puts the Hα stack’s hydrogen into red, where that light already belongs, and keeps the image natural color.
Do not average (halves the nebula, dims stars) and do not add (doubles the sky).
Blend Hα into red only where it leads, above the noise, and only part of that lead.
Balance the RGB on stars first, and combine linear data, never stretched.

Adding Hα to RGB: Natural Color With a Deeper Nebula