Author Topic: Let's start at the very begining... (well the processing part anyway)  (Read 4029 times)

0 Members and 1 Guest are viewing this topic.

MarkS

  • O. A. S.
  • Galaxy Cluster
  • *
  • Posts: 7447
Quote from: Ivor
When you do this what do you have as a background a pure white screen?

That's right - a plain white screen

Quote from: Ivor
The other bit I’m still unclear on is the well depth and whether it is either half full or 2/3 full and more to the point the histogram is a bell curve so where does the well come into it?

Well depth is how many electrons one pixel on the CCD can hold.

I find the histograms displayed by a DSLR quite confusing.  So I simply choose a low ISO and adjust the exposure so that some parts of the image just begins to saturate.  Then halve the exposure and use that halved exposure for creating the flats.

MarkS

  • O. A. S.
  • Galaxy Cluster
  • *
  • Posts: 7447
Quote from: mickw
So having a grot splattered primary, secondary or front element will only degrade the image rather than add bunnies ?

Precisely!

Ivor

  • Past Chairman
  • O. A. S.
  • Star Class
  • *
  • Posts: 595
Another question springs to mind

On the basis

Bias = read noise.
Dark = read noise and exposure noise
Flat = illumination noise.
light = read, exposure & illumination noise and object

Therefore Object =  Light - (Dark + Flat)


Why do you need the bias?

Carole

  • O. A. S.
  • Galaxy Cluster
  • *
  • Posts: 8731
    • Carole's images
Quote
I can see if you step away from a thread for a while it quickly moves on.
I can agree with Ivor on that, have been out all day.

Thanks for the analysis Mark, I have now done what you said and rotated the flat and although it is fairly close, it does not actually hit the same spot, so damn it, the dust must have moved.  I hope you can see with my less mathematical mind why I came to the conclusion that it was due to rotation.

However I find the idea that the dust particle had moved almost incomprehensible since the flattener was covering the DSLR and apart from rotating there was no other movement or air movement around the sensor.  Still at least if it moves then there is a chance I can blow it off altogether with my squeezy airball (whatever it is called). 

Ivor you have started a very interesting discussion here and hopefully we are also answering your questions and you have learnt a bit more than you had intended.

Carole


MarkS

  • O. A. S.
  • Galaxy Cluster
  • *
  • Posts: 7447
Quote from: Ivor
Another question springs to mind

On the basis

Bias = read noise.
Dark = read noise and exposure noise
Flat = illumination noise.
light = read, exposure & illumination noise and object

Therefore Object =  Light - (Dark + Flat)

Why do you need the bias?

Good question.
Definitions of Flats and Darks can differ.  But let's take the definition where the Flats and Darks are what comes out of the camera without further adjustments i.e. without the bias subtracted.  Then the Lights, Flats and Darks all contain bias.

So,
Light = (Object x (Flat - Bias)) + Dark
Hence,
Object = (Light - Dark)/(Flat - Bias)

Mark

MarkS

  • O. A. S.
  • Galaxy Cluster
  • *
  • Posts: 7447
Quote from: Carole
However I find the idea that the dust particle had moved almost incomprehensible since the flattener was covering the DSLR and apart from rotating there was no other movement or air movement around the sensor. 

The CCD cleaning cycle takes place every time you switch the camera on.  It vibrates the sensor at an ultra-sonic frequency in order to shake off the dust. Sometimes it will only move the dust around, instead.  For astro-photography this feature really must be switched off.  Are you sure you have it switched off?

Mark

Rocket Pooch

  • Non-OAS
  • Galaxy Class
  • *
  • Posts: 5027

Ivor

  • Past Chairman
  • O. A. S.
  • Star Class
  • *
  • Posts: 595
Based on what you've said I pondered some more, I'm going to risk it all and challenge your equation (I think this could be suicidal based on this thread and having only been here 5 mins)

Quote
So,
Light = (Object x (Flat - Bias)) + Dark
Hence,
Object = (Light - Dark)/(Flat - Bias)


The purpose of the bias frame is to capture the read noise, from what I understand the read noise is produced during the signal conversion from analogue to digital. Therefore the read noise must exist in all frames and won't increase with length of exposure. It might however increase with time as the CCD ages.

As the frames we take are meant to enable us to eliminate the different types of noise so the different frames must contain the following:
Bias frame [BF] = read noise.
Dark frame [DF] = read and exposure noise.
Flat frame [FF] = read and illumination noise.
Light frame [LF] = read, exposure & illumination noise and object
The goal is to eliminate the read, exposure and illumination noise from the light frames at the same time we try to reduce the signal to noise ratio by taking multiple frames.
So
Master Bias [MB] = ∑ BF
Master Dark [MD] = ∑ DF - MB
Master Flat [MF] = ∑ FF -MB
Master Light [MB] = ∑Object + MD + MF + MB
Therefore
∑Object = MB – (MD + MF + MB)

Discuss...


Rocket Pooch

  • Non-OAS
  • Galaxy Class
  • *
  • Posts: 5027
Off you go Ivor and Mark :-)


MarkS

  • O. A. S.
  • Galaxy Cluster
  • *
  • Posts: 7447
Quote from: Ivor
Discuss ...

Firstly I'll correct your typo - I think you meant:
Master Light [ML] = ∑Object + MD + MF + MB
Therefore
∑Object = ML – (MD + MF + MB)

However, the master flat is multiplicative.  It corrects for vignetting, differential pixel sensitivity and for dust bunnies.  All these are multiplicative effects.  For instance, in the case of serious vignetting, the light intensity at the corners of an image might be 0.5x the intensity at the image centre.  Similarly, a certain pixel might be 1.1x as sensitive as its neighbour and similarly a dust bunny may reduce the light intensity by 0.7x

So, using your notation,

Master Bias [MB] = ∑ BF
Master Dark [MD] = ∑ DF - MB
Master Flat [MF] = ∑ FF - MB
Master Light [ML] = ∑Object x MF/constant  + MD + MB

The constant is chosen to scale most of the MF pixel values into the range 0 - 1.0

Therefore
∑Object = (ML – (MD + MB) ) / MF x constant

Mark

MarkS

  • O. A. S.
  • Galaxy Cluster
  • *
  • Posts: 7447
Quote from: Rocket Pooch
Off you go Ivor and Mark :-)

I've clearly got too much spare time on my hands  :evil:

Fay

  • O. A. S.
  • Galaxy Cluster
  • *
  • Posts: 9210
    • Faysastroimages
Chris, are you critiscising my contribution to the thread?


Mark, you have really warmed to your subject  :D
It is healthier to be mutton dressed as lamb, than mutton dressed as mutton!

Carole

  • O. A. S.
  • Galaxy Cluster
  • *
  • Posts: 8731
    • Carole's images
Looking at the technical conversation between Ivor and Mark I think it is time to duck out of this thread and leave them to it.

Carole

Mac

  • O. A. S.
  • Galaxy Class
  • *
  • Posts: 3360
    • Espania.
getting back to the histogram on the back of your camera.



As you look at it from left to right.

the very left hand side is the colour black,
the very right hand side is the colour white
on a grey scale, each channel will have its own histogram, the one on the back of the camer is
normally an averaged out grey scale of your image.

the height of the graph is the amount of that particular colour.
so looking at the histogram you can guess that the image hame more dark colours then light colours.

If you look at this one


you can clearly see the explanations for over exposing and underexposing and how the appear on the hisogram.
if you look at the overexposed one you can see that the graph is over to the right hand side and you have lost all the detail in the bright parts,
If this was your flat then you would loose all the flat detail in this area.
If you look at the OK one, again there is some over exposure on the right as the graph goes up to the right hand side.

wha you want is for the graph to go about 2/3 across the screen so that you dont loose any detail.

If you are doing a flat image then you will probably have a graph like this.


Mac.
Never, ever, argue with an idiot. They'll drag you down to their level and beat you with experience.
If you argue with an idiot, there are two idiots.

MarkS

  • O. A. S.
  • Galaxy Cluster
  • *
  • Posts: 7447
Quote
what you want is for the graph to go about 2/3 across the screen so that you dont loose any detail.

If you are doing a flat image then you will probably have a graph like this.

I strongly disagree - if you are talking about the histogram on the back of the camera. Here's why:

I've just done an experiment with the histogram on the back of my Canon 400D and with the highlight display switched on (which flashes the over saturated pixels) pointing at at flat white screen.  At 1/20sec no pixels are saturated but at 1/15sec many pixels are saturated. So I halve the 1/20sec exposure time further to 1/40sec to get a decent flat. This works and gives me the correct exposure for a flat.

But when I now look at the histogram on the back of the camera for 1/40sec, ALL the histogram is in the extreme right hand 20% of the chart.  If I selected the exposure time so the histogram was 2/3 along or 1/2 along then I would be severly underexposing my flats.  That's what I mean when I say I don't know how to interpret the histogram on the camera.  It is demonstrably not a linear scale.

For this reason, my advice is to never use the camera histogram to determine the correct exposure for flats.  There is a lot of bad advice on astro-forums concerning this issue.

Mark

P.S. I should also have said that I'm using my slowest ISO for flats.
« Last Edit: Nov 08, 2011, 07:33:39 by MarkS »