Shooting a Mosaic: When One Frame Is Not Wide Enough

Sooner or later you point a telescope at something that will not fit. The North America Nebula spans four degrees; a 500 mm refractor with an APS-C sensor gives you about two and a half. The Veil is worse. You can buy a shorter scope, or you can shoot the object in pieces and join them afterwards.

Joining them is the easy part. Shooting them so they can be joined is the hard part, and almost every failed mosaic failed at the telescope rather than at the computer.

Overlap is not optional

Two panels have to share sky. Touching edges is not enough.

The stitcher works out where each panel belongs by matching the stars it can see in both. With no shared stars there is no way to know how the panels relate, and no software can invent that information. Two panels that merely abut are two separate images.

Plan for 15 to 20% overlap on every edge. That sounds wasteful, but three things eat into it:

• Field rotation and drift. Over an hour on a panel your framing moves. The overlap you planned is the overlap you had at the start.
• Optical distortion at the corners. The outer few percent of most fields is the part you least want to trust as the join.
• Vignetting. Panel edges are the darkest part of a frame, and a seam made from two dark edges shows as a dark band whatever the blending does.

Below about 10% overlap the star matching starts to fail intermittently. It works on one pair and not the next, which is the most confusing way for it to fail.

Shoot each panel as a complete session

A panel is a finished image before it is a mosaic tile, with its own lights, its own calibration and its own stack.

It is tempting to throw every frame from every panel into one stack and let the software sort it out. It cannot. Stacking aligns frames on their shared stars, so frames that only partly overlap will align on the stars they have in common and smear everything else into arcs. You get a mess with a sharp middle, not a mosaic.

So stack panel one, stack panel two, then stitch the results. In Akastroid that is Deep Sky for each panel, then Mosaic for the join.

Match the panels at the telescope, not afterwards

Shoot every panel with:

• The same exposure length and gain. Different exposures mean different noise and different star sizes.
• The same filter. Obvious, and still a mistake people make on a two-night mosaic.
• Roughly the same total integration. A panel with half the time is visibly noisier, and no seam blending hides a noise difference across a join.
• Roughly the same sky conditions. A panel shot through high cloud or with the Moon up sits at a different sky level, has softer stars, and shows as a rectangle.

Stitching can correct the level difference between panels. Akastroid measures each panel’s background and matches it to the reference, which stops the join reading as a step in brightness. It cannot correct a difference in noise, star size or transparency. Those are decided when you press the shutter.

What the stitch does

Three steps, in this order:

1. Placement. Stars in each panel are matched against the reference panel, which gives the offset and rotation to put every panel on one canvas. A panel that overlaps nothing in the main group is reported rather than forced into place, since the right response to “these two do not touch” is to say so.
2. Level matching. Each panel’s background is brought to the reference’s, so no panel is a brighter or darker patch.
3. Feathering. Instead of a hard edge, the panels cross-fade over a ramp. A wider ramp hides a seam better and costs a little detail either side of it. The default is about 12% of a panel, which is invisible on most data.

Planning the grid

Two by two is a sensible first mosaic: four panels, four times the integration, a field twice as wide in each direction.

Work out your field of view first:

field (arcmin) = 3438 × sensor size (mm) / focal length (mm)

An APS-C sensor is 23.5 × 15.6 mm. At 500 mm that is 162 × 107 arcminutes, or 2.7 × 1.8 degrees. For four degrees of nebula with 20% overlap you need two panels across, and the arithmetic tells you that before you waste a clear night finding out.

Then decide the panel centers and enter them into whatever you use to slew. Most planetarium software will draw a mosaic grid for you. The numbers matter more than the tool.

Two by two is four nights, realistically

This part rarely gets mentioned. A mosaic multiplies your integration by the number of panels. Four panels at three hours each is twelve hours for one image, and every panel needs as much time as the others, so you cannot stop early on one and make it up later.

If you get two clear nights a month, a four-panel mosaic is a two-month project. That is fine if you commit to it knowingly, and miserable if you discover it halfway through.

When not to bother

• The object fits. If it fills less than about a third of your frame, you have framing to spare and a mosaic buys nothing.
• You have not filled one panel with signal yet. A noisy mosaic is four noisy images in a wider arrangement. Go deep first.
• Your mount will not repeat. If slewing back to a panel center lands you somewhere different each time, your overlap is a guess.

In Akastroid

Stack each panel in Deep Sky and export or keep the result. Then open Mosaic and add two or more panels; it places them by their stars, matches the levels and feathers the seams. The first panel you add is the reference and the others are matched to it, so make it the one shot in the best conditions.

The stitched field arrives as an ordinary session with the same styles, histogram handles and export as anything else.

Mosaic stitching is included in the free copy, as is every other feature.