FLOAT
EN FR

The path · 2 Intermediate

The full processing: stars apart

You open the project of the first tutorial and add two blocks to its workflow, Denoise and Stars/background: the run computes only what follows them, about a minute. In Post you quieten the stars, thin the tiny ones and give the starless image its local contrast, then you export.

Table 1.The tutorial at a glance.
DurationAbout 10 min, of which 1 min of computing, measured on a PC with an RTX 4060 Ti graphics card (16 GB)
Requirements
  • FLOAT installed
  • The tutorial My first image done: its finished project (1.1 GB, with its cache of calibrated frames beside it) is the starting point
  • Disk space: a few GB more for the results of the two new blocks
Data set Rosette Nebula (NGC 2237) · Sony A7S, 15 s at ISO 4000 · 130/900 Newtonian · 436 lights (1.8 h), 48 darks, 62 flats, the authors of FLOAT; size 6.9 GB. The frames will be offered for download; the link will be given here
The Rosette Nebula after the full processing: Denoise, the stars apart, and local contrast on the starless image. The noise measured on the sky goes from 5.2 to 1.1 levels.
The result. The Rosette Nebula after the full processing: Denoise, the stars apart, and local contrast on the starless image. The noise measured on the sky goes from 5.2 to 1.1 levels.

1Open the project

On the Home screen click Open project… and choose the finished project of the first tutorial, or a copy of its folder. The project opens up to date: nothing is to be computed.

Expected result. The status bar reads all up to date, and the Run button is greyed.

To avoid. The project is saved in its own folder: work on a copy of it if you want to keep the result of the first tutorial as it was.

Open the project
Figure 1.Open the project.

2Diagnosis in Post

Click Post. The layer is the result of the first tutorial. Before any gesture, look at what is right and what is wrong. Right, and to leave alone: the colours, the sky level and the crop set at level 1. Wrong: three points, each with the block or the gesture that corrects it.

Expected result. Level 1’s layer is there, Sky reads −1.5 and Faint tones +1.0. Noise smoothing and the star sliders are greyed: they wait for the blocks that give them something to act on.

Diagnosis in Post
Figure 2.Diagnosis in Post.
The diagnosis before touching Post.
What is wrongHow it is readWhat corrects it
At 100 % the sky is grainy.Zoom to 100 % (Ctrl and the mouse wheel over the picture, or the Zoom menu) on the dark sky in the upper part of the picture: a fine coloured speckle lies on the sky and on the faint gas.Add the Denoise block. (step 3)
The stars are many and crowd the ring.Look at the whole picture, then at the cluster at the heart of the ring: the stars cover the gas, and a swarm of tiny stars fills the sky between them.Add Stars/background, then lower Strength and Small stars. (step 4)
The folds of the ring and the dark globules lack relief.Zoom to 100 % on the dark globules above the centre of the ring: they melt into the gas around them instead of standing out.Add Local contrast on the starless image. (step 12)

3Add Denoise

Click Graph. In the palette, double-click Denoise: the block takes its place in the chain by itself, between Gradient and Color, with the method Pretrained network. It needs no setting.

Why. At level 1, lifting the faint tones lifted the grain too. Denoise removes the noise from the stack before anything is stretched, so the faint tones can then be shown without grain. It is a block of the chain and not a filter of Post: that is why it removes the grain without blurring the stretched picture.

Expected result. The status bar reads to run: ≤ 1 min.

Add Denoise
Figure 3.Add Denoise.

4Add Stars/background

Double-click Stars/background in the palette; it was greyed until Denoise existed. Its method is Neural star removal, with no setting. A Combine layers block then appears by itself in the Composition area, wired to Post.

Why. It splits the picture into two images, the starless gas and the stars alone. From now on each can be treated without harming the other: the stars no longer take the gas’s contrast, and the gas no longer takes the stars’ size.

Expected result. The Run button now reads 1 min: Denoise and Stars/background are to be computed.

Add Stars/background
Figure 4.Add Stars/background.

5Run

Click Run. The Run project window lists every block: Calibration to Gradient, and Color, read up to date and are not computed again; only Denoise and Stars/background are.

Why. This is the way FLOAT works: each block is computed once, and a change only runs again the blocks that follow it. The first tutorial’s run took 35 min; this one takes about 1 min.

To avoid. Before a run, look at the graphics card if another program computes on it: the run then takes longer. The times on this page were measured on an RTX 4060 Ti that nothing else was using.

Run
Figure 5.Run.

6Only what follows was computed

The window ends with 2 blocks computed in about 1 min on an RTX 4060 Ti graphics card (16 GB). The run’s log gives the stages: Denoise 0.2 min, the output 0.2 min and the removal of the stars 0.6 min. Calibration, registration and stacking are not in the list: their results were kept.

Expected result. The status bar reads all up to date and every block of the graph carries its green bar.

Only what follows was computed
Figure 6.Only what follows was computed.

7The new layer

Close the window and click Post. One layer, Color minimal · Combine layers, now replaces the plain one, with two rows under it, Stars and Starless image. The star sliders, Strength, Small stars, Star size and Star color, are active. The picture looks paler: the layer starts from zero, and the Sky and Faint tones of level 1 are not carried over.

Why. The stars are now a layer of their own, with their own sliders: each part of the picture is treated without touching the other. Post builds this layer by itself once the Stars/background block is computed.

To avoid. The layer keeps the template’s name, Color minimal, although the workflow now has Denoise and Stars/background.

A star field at 100 % after Denoise, on the plain layer.
Before. A star field at 100 % after Denoise, on the plain layer.
The same field with the stars separated and recombined: they are smaller and quieter, and the gas reads better.
After. The same field with the stars separated and recombined: they are smaller and quieter, and the gas reads better.
The new layer
Figure 7.The new layer.

8Sky and Faint tones again

In the Starless image group of the layer’s properties, drag Sky to the left, to about −1.5, and Faint tones to the right, to about +1. They are the goals of the first tutorial, on the new layer. Wait until the picture is drawn again after each.

Why. The sky stays a dark, neutral grey, never black: a black sky swallows the faint edges of the nebula, and a light one looks like fog. The faint tones are the extensions of the nebula, what makes the Rosette a rose and not a disc.

Expected result. The two sliders read −1.5 and +1.0, and the picture is back to the dark sky and the faint filaments of level 1.

Sky and Faint tones again
Figure 8.Sky and Faint tones again.

9What Denoise did

Zoom to 100 % on the same dark sky as in the diagnosis, then on the filament at the left of the ring. The coloured grain of the sky is gone, the faint glow of the outskirts shows, and the filament is smooth. The noise measured on the sky of the exported picture goes from 5.2 to 1.1 levels.

Expected result. The same sliders give the same picture, without the grain.

To avoid. Denoise is for pictures, not for measurements: it slightly brightens the faint stars (+8 to +11 % on the bench).

The sky at 100 % at the end of level 1: a coloured grain.
Sky before. The sky at 100 % at the end of level 1: a coloured grain.
The same sky after Denoise: the grain is gone and the faint glow of the outskirts appears.
Sky after. The same sky after Denoise: the grain is gone and the faint glow of the outskirts appears.
The filament at the left of the ring, before.
Filament before. The filament at the left of the ring, before.
The same filament after Denoise: it is smooth.
Filament after. The same filament after Denoise: it is smooth.
What Denoise did
Figure 9.What Denoise did.

10Stars quieter: Strength

With the layer’s top row selected, drag Strength, in the Stars group, a little to the left until the stars over the ring take less light: the cluster stays but stops covering the gas, here about −0.5.

Why. In this field the stars outnumber everything, and the subject is the gas.

To avoid. Go gently: a field without stars looks fake.

The cluster at 100 % before.
Before. The cluster at 100 % before.
Strength at about −0.5: the stars take less light and the gas shows.
Right setting. Strength at about −0.5: the stars take less light and the gas shows.
Strength at about −1.7: the bright stars become dull dots and the cluster looks faded.
Pushed too far. Strength at about −1.7: the bright stars become dull dots and the cluster looks faded.
Stars quieter: Strength
Figure 10.Stars quieter: Strength.

11Fewer tiny stars: Small stars

Drag Small stars to the left until the swarm of tiny stars thins out and the faint gas between them shows, here about −1.

Why. The small stars are the ones that make a picture busy. Taking a share of them away is what most imagers do with a star-rich field. Small stars and Strength are independent: their order does not matter.

To avoid. Do not push it to the far left: grey round dots appear in the sky where the stars were taken away, and the field looks empty.

A star field at 100 % before.
Before. A star field at 100 % before.
Small stars at about −1: the swarm thins out and the faint gas shows.
Right setting. Small stars at about −1: the swarm thins out and the faint gas shows.
Small stars at the far left: faint grey round dots appear where the stars were, and the field looks empty.
Pushed too far. Small stars at the far left: faint grey round dots appear where the stars were, and the field looks empty.
Fewer tiny stars: Small stars
Figure 11.Fewer tiny stars: Small stars.

12Local contrast on the starless image

Click the Starless image row, then Filter, Contrast and Local contrast. Drag its Strength to the right until the dark globules and the folds of the ring stand out from the gas around them, here about 0.5. Leave Scale at its default, 40 px.

Why. The filter goes on the starless image and not on everything: over the stars, the same filter makes them hard, square and haloed. That is the reason the stars were separated.

Expected result. The filter appears under the Starless image row, and the stars are untouched.

To avoid. A Scale of 60 instead of 40 shows no visible difference (0.7 level on average): leave it alone.

The globules at 100 % before.
Before. The globules at 100 % before.
Strength at about 0.5: the globules and the folds stand out.
Right setting. Strength at about 0.5: the globules and the folds stand out.
Strength at about 1.2: the globules turn black with dark haloes and the gas becomes blotchy.
Pushed too far. Strength at about 1.2: the globules turn black with dark haloes and the gas becomes blotchy.
The filter on the Starless image row: the stars of the cluster stay soft.
Right place. The filter on the Starless image row: the stars of the cluster stay soft.
The same filter over everything, stars included: the stars turn hard, square and haloed.
Wrong place. The same filter over everything, stars included: the stars turn hard, square and haloed.
Local contrast on the starless image
Figure 12.Local contrast on the starless image.

13Export

Click Export…, choose JPEG in Format and click Export. The file is named after the project, in its export folder.

Export
Figure 13.Export.

14The image is written

The window says how many files were written: the image and its processing history. Close it. Press Ctrl+S to keep the project with its two new blocks, its new layer and the three settings.

The image is written
Figure 14.The image is written.

Blocks used