Original subtitles

Hi, in this video I'm going to show you

how you can take a stacked

astrophotography image of Andromeda and

in just seven steps process it to bring

out the detail. You don't need any paid

software or any kind of fancy AI tools.

You can download the data for free from

my GitHub page. It's just simple,

minimal processing that will help you

get a nice image. This video is aimed at

those who are starting out in

astrophotography, but I think there

might be useful information for anyone.

Welcome to Deep Sky Detail.

So, I've got serial opened and I'm

assuming that you already know how to

stack an astrophotography image. So, you

take multiple images and then stack

them. If you don't, I've got a video on

how you can stack subframes.

But for now, we're going to assume that

you already have your stacked image and

you just want to process it using very

simple steps to get the most out of what

you've worked on. So, I've got Serial

open. Serial is a great piece of

software that's dedicated to processing

astrophotography images. We'll also be

using is a free Photoshop

clone that's very powerful. These two

programs are going to be able to help

bring the most out of your image. So,

let's go ahead and open up the image

that we want. Click open up here, this

button, and then find the image on your

computer. Now, what's going to pop up is

your astrophotography image in its just

raw state. It's nothing has been done to

it. All of the data are there. It's just

waiting to be brought out with the

software. A very quick way to see the

data is to go down to here to this

button, click this button that says

linear, and then change it from linear

to auto stretch. Now, we can do better

than this auto stretch. So, we're just

using this so we can see what we're

doing as we process it. This auto

streretch has not actually been applied

to the image. It's just us previewing

what the image could look like. And it's

just there so we can see what we're

doing. So, step number one is to crop

the image. So, take your cursor,

leftclick somewhere where there's no

funky stacking artifacts. So all this

green, magenta, and all of this yucky

stuff up in this corner here, those are

all stacking artifacts. We want to get

rid of that. We want to crop that out.

Take your cursor, click where there's no

stacking artifacts, and then drag it to

what we want to keep, zoom in, and start

to exclude what we don't want to keep.

So when the cursor changes to these

arrows, then you can click on it and and

change the selection. Once we've gotten

the selection that we want, right click,

click crop, and then crop. And then all

of that stuff will be cropped out and

it'll be no longer part of our image.

Step number two is to do background

extraction. So when you're taking an

image like Andromeda in broadband and in

true color, light pollution and may and

possibly the moon are going to create

gradients along your image that aren't

part of your deep sky object. So this

part is darker, this part is lighter. We

want this to be uniform. So let's go to

image processing

background extraction

and make sure that RBF is selected. RBF

is a very powerful non AI tool, meaning

it's not a neural network. And then what

we need to do is create sample points

around the image. Make sure that add

dither is checked.

We create sample points around the

image.

You don't need a lot of sample points,

just enough to where it can get a good

representation of what the light

pollution looks like in your image.

Making sure that you don't click on the

galaxy. If you do click on the galaxy,

you can right click to remove it. Once

we've got 10 or 15 spread around the

image, we can click compute background.

And already all those a lot of those

gradients are gone. Very powerful tool.

Let's go ahead and click apply to save

those changes. Next step we want to do

is deconvolution. So let's go to image

processing

filters and then deconvolution. Now what

deconvolution is going to do is help

sharpen the image. So the first thing we

want to do is generate a point spread

function. A point spread function tells

serial how the image is blurred. So we

go click on PSF from stars. Generate

PSF. Make sure that the PSF size is big

enough to make a good representation of

what your stars look like plus some

background. We got to figure out how

sharp we want to make the image. That's

basically controlled by this parameter

here that says iterations. The more

iterations you have, the sharper the

image will be. The fewer iterations you

have, the less sharp. Editor me from the

future here. In the original video, I

should have mentioned that you want to

be using Richardson Lucy deconvolution

along with gradient descent as the

algorithm method. Otherwise, otherwise

88 iterations is going to be way too

much. gradient descent gives us better

control and it's better overall. I know

that for this image 80 is 80 iterations

is pretty good. So I'm going to click on

80 and then click apply. Okay, I just

ran that. It'll take a few minutes.

Let's just zoom in just check to see if

there are any ringing artifacts around

the stars. I don't see it. It's not too

bad. We could click on this undo button

to see what it was like before

and after. And if we zoom in, we can see

that the stars are getting a little bit

tighter. So this is before

and this is after. Looks pretty good to

me. If things are too sharp, again, you

press the undo button, decrease the

iterations. If it's not sharp enough,

press the undo button and increase the

iterations. It's really that simple. It

does take a few minutes to run the

deconvolution. So if instead of

deconvolving the whole image, you just

want to deconvolve part of the image,

zoom in into the galaxy, leftclick and

drag to make a selection and then right

click,

select ROI and set ROI to selection.

And then instead of applying the

deconvolution to the whole image, you

can just apply it to the ROI preview.

When you're done and you're happy with

the amount of iterations you have, then

just do rightclick

ROI and clear ROI. Okay, so

deconvolution is done. Step four is we

want to color calibrate the image. Now

for this image, there's a lot of green

noise in the image, which is not very

good. Stars are generally not supposed

to be green. So let's go to image

processing and then we will go to remove

green noise.

Click apply and suddenly all that green

around the stars is gone. Nice. Let's go

ahead and close this. Now the image

after removing the green noise is

actually pretty well color calibrated.

But if you find that it's not, if it's

still too green or red or blue or

whatever, then you can go to image

processing color calibration. Then click

on color calibration.

Zoom in to part of the background of the

image, one part of the image that

doesn't have any galaxies, stars, or

nebula in it, and select it. Click on

use current selection, and then do

background neutralization. And that's

all you need to do. After that, you can

close out this dialogue box. All right,

this looks pretty good by itself, but I

think we can stretch this a little bit

better. So, instead of using serial

default auto stretch, let's turn this

back to linear. and we'll go to image

processing stretches and let's do first

the arc sign transformation. It's going

to warn us that some of these values may

be clipped. So let's go ahead and

rescale the values and then increase

this stretch factor in the arc sign

transformation tab. Hi editor me again.

I just want to point out that in the

original video I stretched this way too

fast. So you probably don't want to go

all the way up all at once. You might

may want to do an iterative step. So

maybe go like 100 200 and click apply.

You can also change the black point here

as you stretch. But generally like to

mess with the black point after I've

done the inverse hyperbolic stretch

transformation because you can see the

histogram better and know when you are

making the picture too dark. Now what

this is doing is it's stretching the

image in a way that helps preserve more

of the color. So we'll go and click

apply. And now the image is way too

bright. So we need to do a normal

stretch on it, not an arc sign stretch.

So let's go to image processing

stretches.

and then histogram transformation. You

can see that the the peak of the

histogram is way towards the right here.

We want to bring that back towards the

left to make it dark again. So let's

take this dark slider here and move it

this way

all the way until the histogram is on

the left.

Let's go ahead and click apply. If

you're just starting out, you could go

straight into a generalized hyperbolic

stretch or removing the stars and then

stretching things and then putting the

stars back in. But I want this to be a

really simple tutorial, something that's

really easy for beginners to do. And as

you progress in the hobby, you'll be

able to cherrypick more advanced

techniques that you prefer and add them

into your images. So, for this, all I'm

going to do is just click this button

here that performs the auto streretch

onto the image. It'll automatically

stretch your image, and it does an okay

job. Let's go ahead and click apply.

Now, one of the things about the auto

stretch in serial, it's too aggressive

for my tastes. So, I'm going to take

this middle slider and just

move it back to the right a little bit.

And by doing that, it helps darken out

the background and make sure that we're

not overstretching things. Click apply.

And there we go. It's already looking

like a nice image, except there's a

little bit of color in the background

and a lot of little bit of gradients in

the background that we probably want to

get rid of. So, let's go ahead and bring

this into And for step seven,

we'll fix the background. So, let's go

and do this. We'll click on this arrow

button to download it. I'm just going to

download it as a TIFF file because TIFF

files are more easy for to work

with than TITS files. can do fits,

but it can be kind of buggy sometimes.

I'm just going to call LRGB stretched.

Tiff, and I'm going to save it into a

32bit floatingoint image. Okay, now that

I've got open, I'm going to go

ahead and open up the file that I just

created in serial. And a dialogue box

will pop open about sort of the color

space. Just keep is fine. And we've got

our image opened up in And what we

want to do is clean up the background.

And so the easiest way to do this is to

convert this to lab space and then use

the L channel in the lab space as the

luminance. And I I'll just show you what

I mean. With the image selected here,

let's go to colors, components, and

decompose.

A dialogue box will pop up and it'll ask

us how we want to extract the colors.

Now, the default is probably RGB, but

we're going to go to lab and then click

okay. What's going to happen is another

tab is going to pop up in

and it's going to have the luminance

information separated from the color

information. The color information is

actually in two channels, the A channel

and the B channel. The B channel is your

blue and yellow colors, while the A

channel is your red and green colors.

Let's go ahead and take this B channel,

deselect the L channel here, make it

invisible. Take this A channel. We're

going to go to filters,

blur, Gaussian blur, and then

we're going to zoom in to make sure that

we're actually blurring things. We can

go to split view. And we can see that

it's very noisy. This is going to be

color noise. Once we blur it, it goes

away. So, click okay. And then we'll do

the same thing with our B channel. So

select our B channel, deselect the A

channel so it's not visible. And we can

do CtrlF to do the same filter on this.

So Ctrl + F and then this is blurred.

This will help get rid of the color

noise as well. One small trick that we

we might want to do is to stretch just

the luminance channel just a little bit

more. Why we're doing this now in

is that it's not going to be stretching

the color data, which could introduce

noise that can be really distracting.

So, let's go to colors and then levels.

And then we can take this midpoint

slider,

move it up a little bit. Take this black

point slider,

move that down.

And that looks that looks a little bit

nicer than it did before. This is the

after. This is the before. The

background's just a little bit darker.

The galaxy pops a little bit more. Now

we go to colors, components, and

recompose. And what that's going to do

now, take the lab data and put it back

into that first tab. So let's click on

this first tab up here.

And we can look at the before and after.

So this is after we cleaned up the image

a little bit. And this is before.

After.

Before. After. And now the background is

a little bit darker. The galaxy pops a

little bit more. It looks a little bit

nicer. So there's still a lot of

background noise. So like if we zoom in

very far, you can see that there's

little it's a little bit splotchy.

There's a lot of blue in the background,

a lot of yellow in the background. It's

not very even.

So zoom back out. What we're going to

do, go back to this lab tab. Take the L

layer. Press control C if you're on

Windows to copy it or command C on Mac.

go back to our original tab and Ctrl +V

to paste that layer on top of this

layer. Now, we've got the luminance data

here and we've got this color data

underneath it. Let's reorder this layer.

Click the L copy that we just had and

move it down underneath the LRGB data.

Now, select the color layer. We're going

to select this RGB image and change its

mode from normal to LCH color.

It's going to tell that the RGB

layer is where the color information is

and this L layer is where most of the

data is. So, we can actually just rename

this as the color layer and this as the

luminance layer. If we zoom in,

we can see that there's still some color

noise. So, we can click on this color

layer, go to colors, levels, and take

this midpoint slider here and take it to

the right. And as we take it to the

right, the image is going to be

desaturated.

If we take it all the way to the right,

only the brightest parts of the image

will have color. So, let's let's move it

back until we start to see no color in

the background.

Yeah, that looks pretty good.

We can turn on split view

to see the before and after. So, let's

zoom in in the split view. This is after

and this is before. You can see that

there's a lot of color noise in the

background in this one. And there's not

a lot of color noise the background of

this one. And just like that, in seven

easy steps, we have a very nice image.

And all we have to do now is export this

as whatever type of image we want,

whether it's a PNG, a JPEG, another TIFF

file, and it just looks nice. Well, I

hope you enjoyed that tutorial. I wanted

to make something that was very

accessible for beginners to do that

doesn't involve a lot of very complex

steps. I generally find that if you have

a lot of data, so you have so for

example I have 15 hours of total

exposure time for this andromeda image

that you can do very simple processing

and get a very good image without

needing a lot of fancy algorithms. And

if you are new to astrophotography and

you want to know how to stack subframes,

I've got a video on that too using

serial and cerillic. Thanks for

watching.