Log in

View Full Version here: : Flats for DSLRs


Doug
12-08-2007, 12:55 PM
G'day all,
I am wondering what the general method used for generating and using flat field images is amongst those who use them?

I have proved to my satisfaction that the colour temperature of the light source used in flat production effects the eventual colour balance of the converted image.(one-shot Colour, not mono)

No matter what form of light source is used, ie Tshirt, light box, or dome flats etc. there will be a colour component of some degree.
I have tried to grayscale a flat and use that, however it does not seem to work very well.

I have previously used a blue sky and an overcast sky. Both provide an effective flat field though the colour conversion parameters are very different in each case.

I have a light box about 95% completed which will at least give me repeatable colour temp values, but the issue will still remain for those not using a light box.

I have found that using a G2V star image to establish the correct colour conversion parameters for each new flat field works well but it can be a bit tedious.
I will soon post a zip file containing a dark subtracted G2V star image for those who might like to check their colour balancing, but in the meantime I would appreciate getting a consensus of flat field calibration techniques being used.

Cheers,
Doug

Rigel003
12-08-2007, 01:45 PM
I'm just a beginner in this field but have been using t-shirt flats with indirect tungtsen light illumination indoors. This tends to give a pinkish cast to the flats. I queried Mike Unsold about converting them to greyscale before automatic callibration in Images Plus and he responded that it was unnecesary.

Doug
12-08-2007, 09:18 PM
G'day Graeme, that is right, it would be unnecessary.
In fact my tests show that the efficiency of a grayscaled flat is diminished anyway, regardless of any effect on final colour balance.

This whole thing is probably not an issue for those using a constant light source, only those using either the direct sky light or reflected sky light, such as using a T shirt in daylight on field trips.
In Ron Wodaski's book, he suggests pointing the scope at an area under a tree or similar. That is fine for monochrome imagers, but I think it would be an undesirable variable for use with one shot colour imaging.
Lets see what others are doing,
Doug

JohnG
13-08-2007, 10:18 PM
Hi Doug

I still use the old white T-Shirt method, I point the telescope at a clear blue sky as far away from the sun as possible, take some test exposure and check that the Historgram is offset to the left third of the frame and take about 20 or so RAW's. As you have already said, the colour dosen't matter when the Master Flat is made.

I have the camera in the same orientation and focus position as when my Lights are taken although there is another school of thought that advocates rotating the camera 90 degrees after each 5 or so exposures, not sure about that one though :shrug:.

Cheers

mostschaedel
16-08-2007, 02:11 AM
Hi i am doing it the same way. In addition i am rotating the camera together
with the telescope or telelens after 3-5 images. Number of flats
usually 16 to 25.
The Masterflat gets anyhow monochrome so this works nicely to
remove the dust spots and the vignetting of the optical train.

dugnsuz
16-08-2007, 07:23 AM
Hi Doug,
This link may be of some help?
http://www.iceinspace.com.au/index.php?id=63,211,0,0,1,0
Cheers
Doug!!

Doug
19-08-2007, 07:49 PM
Thanks for the link Doug, Eddie's article is focused on Monochrome CCD imaging. I'm interested in the effect of colour of flats for colour imaging.

Doug

Doug
19-08-2007, 08:08 PM
G'day John and Gerald, Not too sure of the reason or benefit in rotating the camera.
That alerted me to the effect colour has on the final image was that like you, I took some sky flats using a mid afternoon blue sky, then G2V calibrated my S/W. I needed to take a new set of flats, but did nor wait for a clear blue sky: I collected data from an overcast sky instead. When I applied these flats to their intended light frames, the colour balance was all screwy. Colour returned to normal if I used the older flats, though naturally enough these no longer effectively flattened the image or removed dust motes.

Gerald, when you say "The Masterflat gets anyhow monochrome ", are you referring to the fact that you are applying the flats before colour conversion, or are you converting the flats to monochrome, ie. removing the Bayer matrix coded data? Sorry, just not sure of what you mean.

cheers,

Doug

[1ponders]
19-08-2007, 08:14 PM
Doug, I've found that using Eddies design is fine for a 300D. You collect all three channels when you take the flat anyway. The red will be stronger but it won't matter too much. One thing I did do that Eddies didn't show was to put a dimmer switch inline and wire the bulbs in parallel and not in series.

[1ponders]
19-08-2007, 08:19 PM
Check out the link in this thread
http://www.iceinspace.com.au/forum/showthread.php?t=2556&highlight=flats

and check out the discussion in this thread
http://www.iceinspace.com.au/forum/showthread.php?t=19263&highlight=box

JohnG
20-08-2007, 09:47 AM
Hi Doug

I have never really understood why the camera needs to be rotated as well, the most important thing is the focal position and orientation to match your Lights. I mentioned this only because I have seen it come up in various discussions before.

I only take Flats of the blue sky opposite the sun, I would doubt very much that using a cloudy sky would be of any use, the different colours and textures in the clouds would completely unbalance the final result and, in my opinion, render the Flats useless.

When I get back into imaging next year after I move, I might give it a go just to see.

Cheers

[1ponders]
20-08-2007, 10:38 AM
Just to clarify, are we talking about rotating the camera itself between batches of flats or rotating the light souce in relation to the camera and lens system. If you rotate your camera then you are defeating the purpose of taking flats.

JohnG
20-08-2007, 02:01 PM
Hi Paul

I am talking about rotating the camera itself, I agree, you do, in my opinion, render the Flats useless.

Just to clarify, this method has been promoted on a number of Groups and was, infact, mentioned to me, by a very well know photographer on this group some time ago. (No names, no packdrill) :P

I have chosen not to use it.

Cheers

Doug
20-08-2007, 07:19 PM
I have heard of rotating the imager when imaging light frames to cancell out dust motes, but even so, a flat to get rid of vignetting would be marginal at best.

Doug
20-08-2007, 08:10 PM
John, you might be right about the overcast sky. I used one that was as even as my eyes could tell, and felt that since the FOV would be relatively small I should get away with it. When I did use the overcast sky flats, as flats they did work well, it is just that I had to change the colour scaling values to regain G2V balance. The light frames were however flattened effectively.
cheers,
Doug

[1ponders]
20-08-2007, 08:38 PM
Doug if you have dust motes on you glass cover of the camera chip then rotating will not make a difference and flat fielding will remove them. If you have them on glass elements outside the the camera then rotating the camera and then dividing the flats into the image will then ADD artifacts to the image. Also properly taken flats should remove nearly all of your vignetting, depending on how bad it is.

It is important to keep you camera in the same orientation for flats as with lights. If we take an extreme example and you are imaging through a telescope that has a rectangular cross section rather than round, if you rotated your camera your flat would have it's dark areas completely out of alignment to your light frames. The same thing applies to a normal telescope except you are trying to counter imperfections in how your lenses/mirrors transmit/reflect light to your chip surface.

You also mustn't change the focuser position. This will change the size of the different optical defects.

Doug
20-08-2007, 08:46 PM
Let me try and re articulate the line of my enquiry.
(I am aware of many light box designs and research into illumination uniformity that dates back to last century in the USA. the cubic box being one of them)
Blue skies are nice, however the 'Blueness' is influenced by the amount of smog/water vapour suspended in the atmosphere from time to time. Actual illumination levels are by and large seen to be irrelevant, only the actual colour temperature is of an uncontrollable nature here.
A light box using any type of incandescent filament lamp will also not give completely stable colour temperature illumination. Varying the applied voltage will vary the operating temperature and hence the colour temperature of the resultant light.
When imaging with a monochrome sensor and using red, green and the blue filters in order to create an RGB colour image, the use of those same colour filters tend to regulate the colour temperature, regardless of the the sub frame type. That is to say that the colour filters regulate the colour temperature of the light being used to produce a light frame and a flat field frame to the same value; therefore as long as the illumination source used is sufficiently broad, any variation within reason will be cancelled out by the filter action.
This does not apply to any one-shot colour sensor. Any variation in the light source will effect the RGB ratios at the sensor; if it did not, we would have no colour.
Therefore my quest is not for info on mono focused American designed light boxes, rather what steps if any are being used by people using one-shot colour imagers to regulate the actual colour temperature of their light sources Colour temperature is not to be confused with light intensity.

So I'm interested in particular in the performance (not design) of light boxes that use LED's,(or even Electroluminescent panels) is there a known variation of colour temperature with variations in forward current flow?
thanks in advance for any info on colour temperature issues as applied to flat field acquisition for one-shot colour imaging.
cheers,
Doug

[1ponders]
20-08-2007, 08:58 PM
You will find that our good mate Dennis is in the process of creating a light box using LEDs. I don't know if it completed yet or if he has used if if completed.

As for using an incandescent system, from memory it doesn't really matter. As long as your illumination is strong enough in each channel to illuminate the optical defects, the flat can be normalized and then divided into the colour image with np. When taking a flat field using Eddies design with a 300D or similar colour chip, as long as my histogram of the image (the filled pixel wells) is between 1/3 and 1/2 the flat field works fine. The flat temperature isn't a problem, and it is strongly in the red.

Doug
20-08-2007, 09:01 PM
emphasis added.
Paul, Dust motes are not unique to flat field frames. If one were to take several light frames, each rotated, then when those light frames are combined using a median or Sigma or Sigma clip type algorithm any outlier pixels will be eliminated...cancelled out. This will include cosmic ray hits, dust motes, hot pixels meteor trails, mostly satellite trails aircraft light trails etc.
I agree it will not do a lot for flat fielding images; I think the idea is to avoid the need for flats, which if vignetting is not an issue will work.

Doug

[1ponders]
20-08-2007, 09:20 PM
Sorry misread the original. Yes if you rotate your camera when imaging then you may eliminate some of your dust motes. It would depending on how you combined them, but your mote shadows may not be considered as outlier pixels. It would depend on what they were lying on when rotated, both within the image and on glass surfaces. If they were on the filter or reducer optics then the motes would still appear in the same spot of the image no matter how you turned your camera.

It is very hard for vignetting not to be an issue. It can come from a number of sources, one that is outside the effects of shadowing by internal structures is light falloff (which technically isn't really vignetting which refers to shadowing). Unless your lense can produce a perfectly flat image then the area being illuminated by your lens/mirror will nearly always have an area of uneven illumination or light gradient of some sort. The best way to deal with this type of gradient is with a flat. You can sometimes remove it using an Antivignetting technique in Photoshop or similar, but doing this throws away information. It would be much better to remove as much of the the problem before you worked on it in PS

Doug
20-08-2007, 09:49 PM
Paul, you are wrong in your reasoning re light frame artifacts because even if the artifact stays in a fixed position relative to the imaging sensor, they will cerainly not remain fixed relative to the image. When 2 or more points (stars) are aligned any pixel that is at odds with those belonging to the image will be outlier pixels and hence eliminated. This will apply in small part to vignetting, but is not as good as a flat field would be.....admittedly.
Can we keep this thread on topic please?

Doug

[1ponders]
20-08-2007, 09:54 PM
Sure.

mostschaedel
21-08-2007, 06:56 AM
Hi Doug!
I am using IMAGEPLUS for the flatfieldcalculation. Imageplus automatically
converts the flatfield before division to monochrome. Using bluesky
flatfields without conversion to monochrome would create uncorrectable
color errors!

Concerning camera rotation:
Camera rotation alone DOES NOT MAKE SENSE.:scared:
I do Camera rotation together with the OPTIC to reduce the minimal gradient
which could be there in the blue sky 90 degree away from the sun.

Doug
22-08-2007, 11:21 PM
Aha!! Thank you Gerald! A bit of light begins to dawn. A lot of people use Images Plus, whereas I do not. If Images Plus converts the flat fields to Gray-scale, that would explain why nobody is plagued with colour variability from Flat series to Flat series. MaxDslr does not do this as far as I can understand Canadian English. Leaving the Flat Field frame with the colour encoding will make the calibrated image influenced by the colour balance of the Flat. Now I begin to understand why people are having trouble understanding where I'm coming from here.

I have tried to grey scale a flat, but found that the process reduces the effectiveness of the flattening. I will need to re expose some flats at a higher density, so that after grayscaling they will have a 30%-50% intensity.
Nevertheless, I think the ideal will be to get a light source that will give a fairly constant colour temperature or spend money on Images plus....I think a steady lightsource will be cheaper. I agree with your comments on rotating camera only.
cheers,
Doug

Paul I will try to reply to you tomorrownight

JohnG
23-08-2007, 09:50 AM
I suppose I should have said I have been using ImagesPlus for quite some years now for all my DSLR work.

Just to be clear, I do not rotate the camera or advocate rotating the camera when taking Flats.

Cheers

Phil
23-08-2007, 07:51 PM
This is what i made to make my flats. It has 4xwhite LEDs that can be dimed inside. The LEDs are defussed with the white defusser that goes in front of the telescope. This works great and is very cheap to make.
Phil

Doug
24-08-2007, 07:40 PM
What processing program do you use Phil?

Doug
24-08-2007, 07:45 PM
I picked up a worthwhile idea from Chile the other night.
http://www.astrosurf.com/antilhue/light_box_for_ccd_flats.htm

Daniel Verschatse is a very accomplished Astro-imager with considerable experience and many fine images to his credit.
I intend using some white LEDs an combo with some yellower tungsten lights, though for the DSLR work any light will do as long as I can get constant colour temperature for use with MaxDslr.

cheers,
Doug

Phil
27-08-2007, 08:07 AM
I use Maxlm DL for all my camera control and and processing my darks flat etc. Then i use photoshop.
Phil