Why Your Nebula Came Out Red, Brown or Washed Out

You stack your first serious set of frames, stretch it, and the Orion Nebula appears — floating in a brown-red haze that fills the entire image. The nebula itself looks flat. The stars have no colour to speak of. Somewhere in the corner the background is noticeably brighter than the opposite corner.

This is not a beginner mistake in the sense of something you did wrong at the telescope. It is what correctly-captured data looks like before three specific corrections, and it happens to nearly everyone. What follows is how to tell which of the three you are looking at, because the fixes are different and applying the wrong one makes things worse.

Cause one: the sky is not grey

Look at the histogram of your stacked image, with the red, green and blue channels drawn separately. If the red trace sits above the green and blue across the whole range — not just at one end — you are looking at a colour cast rather than a colour.

There are two independent reasons for this, and they need different corrections.

An additive offset. Light pollution is not white. Sodium and older LED street lighting push hard into the orange and red, and that adds a constant to your red channel everywhere in the frame. It raises the floor.

A multiplicative imbalance. Your camera’s colour channels do not have equal sensitivity, and neither do the filters in front of them. A one-shot-colour sensor typically has twice as many green photosites as red or blue, and the response curves differ. This scales the whole channel, not just the floor.

The distinction matters because subtracting a constant fixes the first and does nothing about the second. If you neutralise the background and the image still looks red in the bright regions, you have corrected the offset and left the gain.

For the offset, the correction is a per-channel shift, not a scale. Shifting moves the sky to a common level and leaves every difference above it exactly as it was, so the nebula keeps its colour while the sky loses its tint. Scaling would drain the object along with the background — which is why “just reduce the red channel” produces a grey, lifeless nebula.

For the gain, you need a reference for what neutral actually means. Averaging the stars towards grey is the usual heuristic and it is often close, but it assumes the field’s stellar population is average, which some fields are not. The rigorous version matches your field against a star catalogue with measured colours and fits the channel gains that reproduce them — a measurement instead of an assumption, and it reports a residual so you can tell whether it worked.

One caution that catches people out: if you white balance against stars in a field like Orion, the stars are sitting on bright red nebulosity. Measure each star against a single sky level for the whole frame and you count that nebulosity as part of the star’s flux, which reports the nebula’s colour as the camera’s colour. The white balance then corrects for a bias that was never in the sensor, and pulls the red out of your nebula. Each star’s sky has to be measured from an annulus immediately around that star.

Cause two: the gradient

If one side of the frame is brighter than the other, or there is a broad dome of light in a corner, that is a gradient. Its sources are light pollution from a specific direction, the moon, twilight, or vignetting your flats did not fully correct.

Gradients matter more than they look, because stretching amplifies them. A gradient that is barely visible in the linear data becomes a dominant feature once you have pulled the faint end up. Worse, it corrupts the black point: the stretch places your sky at a target level based on the image’s median, and a gradient means there is no single sky level to place.

The standard correction fits a smooth surface to the background and subtracts it. Two details decide whether that helps or hurts.

The surface must be low order. A first, second or third-degree polynomial can describe a tilt, a dome, or an off-centre dome. That is the right level of flexibility, because light pollution genuinely varies smoothly across a frame. Give the fit more freedom and it starts following your target.

Object regions must be excluded from the fit. This is the step whose absence makes background extraction destructive. A sample taken in the middle of a nebula reports the nebula’s brightness as “the sky level there.” Fit a surface through those samples and the surface rises to meet the object — then subtracting it digs a dark bowl out of the very thing you photographed.

If you have ever run background extraction and ended up with a dark halo surrounding your target, that is exactly what happened. The fix is not a gentler setting. It is excluding the samples that are standing on the object, so that the surface is fitted to sky and only sky.

The correction should also be a subtraction with the mean added back, not a pure subtraction. Removing the slope while keeping the overall level is what you want. Driving the background to zero gives you a black sky, and real sky is never black — a truly black background is one of the clearest signs of over-processing.

Cause three: the sky is simply too bright

Your image can be free of casts and gradients and still look washed out, because the background has been lifted too far.

The black point is usually set by moving the image’s median to a target brightness. In a frame that is 90% empty sky, the median is the sky — so a target chosen for a frame where the object fills half the picture will lift your entire background to mid-grey.

This is worth measuring rather than eyeballing. A typical wide-field frame of Orion is around 98% sky by area. At that ratio the target background needs to come down substantially — on real data I have measured the right value at around 0.057 where the generic default was 0.160. That is nearly a factor of three, and the difference between a photograph and a hazy grey rectangle.

If you are reaching for the exposure slider and pulling it well negative every single time, that is the symptom. The default is not tuned for how much of your frame is sky.

Telling them apart

A quick diagnostic:

• Uniform tint across the whole frame, histogram channels offset from each other → colour cast. Neutralise the background with a per-channel shift, then check whether a gain correction is still needed.
• One region brighter than another, or a dome → gradient. Fit and subtract a low-order surface, with object regions excluded.
• Colour is fine, gradient is gone, still looks hazy and grey → background level. Bring the black point down.

They stack, and they are usually all present at once. Correct them in that order: cast, then gradient, then level. Doing it the other way round means setting a black point against a background that is about to move.

What not to do

Don’t crush the blacks. The temptation, when faced with a washed-out image, is to pull the black point until the background is properly dark. This looks decisive and destroys the faint outer regions of your target, which live just above the sky. A real astrophotograph has a background that is dark grey, not black, and faint structure emerging from it.

Don’t fix colour with saturation. A desaturated nebula caused by a bad white balance does not become correct when you raise saturation — it becomes a more colourful version of the wrong colour. Fix the balance first.

Don’t apply background extraction twice. Each pass fits and subtracts. A second pass on already-corrected data fits mostly noise and object, and starts removing signal.

The order matters, and so does when

One more thing, easy to miss: these corrections belong on linear data, before the nonlinear stretch. A colour cast is a linear property — a constant added to a channel, or a channel scaled. Once you have applied a nonlinear stretch, that constant has been mapped through a curve, and subtracting it afterwards no longer removes it cleanly. Same for the gradient.

The practical version: do calibration, gradient removal and colour correction on the stacked linear master. Stretch after. If your workflow stretches first because that is when you can finally see what you are doing, you are correcting a distorted version of the problem, and the corrections will not quite land.

That single ordering rule fixes more images than any individual slider.