best color space for AcrCalibrator

[DrSaltz]

Hey!

New to the group...

I use Adobe CS2 and ACR 3.7:
Of the several color spaces which is the best to use for camera
calibration?
My nikon d200 gives me the option of aRBG vs sRGB
but give different values for the same file.
My goal is to print quality 19x13 inch pohotographs.

Doc Saltz

[Steve Sprengel]

The camera selection for colorspaces are for JPGs that come from the camera unless you're using a Nikon-specific RAW converter, which Adobe Camera RAW is not. 

 

ACRCalibrator works with RAW format files not JPGs and internally uses ProPhotoRGB as the colorspace during its computations.

[DrSaltz]

Thanks Steve,

ProPhoto it is then, (just when I was ready to go back to adobe RGB.)

By the way, I ran the AcrCalibrator script after opening the NEF file
in ACR, then opening it in CS2, as a sRGB, aRGB and a ProPhoto RGB.
Over all the values were not too far off from each other with the sRGB
and the ppRGB being nearly identical. Maybe the "8 bit file" limited
the variations.
My Canon Pro9000 printer is reputed to have a wider color gamut than
other printers but I am not able to find details on its color space.

Doc Saltz

> Doc Saltz- Hide quoted text -
>
> - Show quoted text -

[Steve Sprengel]

Regardless of an image's internal colorspace (sRGB, aRGB, ppRGB) it should look the same when just opened in Photoshop because Photoshop converts the internal colorspace to the monitor's colorspace for display.  The colorspace of the image does make it look different in IE or FF because those browsers always assume the colorspace is sRGB.  This is why you have to convert to sRGB before posting online.  Safari is the only browser that honors the internal colorspace but most people are using IE or FF.

 

The internal colorspace that ACRCalibrator uses does not dictate the colorspace you should use in your processing.  ACRCalibrator uses the large ProPhotoRGB colorspace to keep the colors as accurate as possible while the sliders are being manipulated instead of the clipping that would occur with a smaller colorspace like AdobeRGB and especially with sRGB.

 

Here is a series of images and captions to illustrate the clipping and subsequent posterization or color-shifting that occurs when an image in sRGB or AdobeRGB is manipulated in Photoshop, as a reason why using ProPhotoRGB is good:

 

http://www.pbase.com/ssprengel/tech_colorprofiles

[MacUser]

Hello Steve,

Thanks for posting the link, I looked at the color profile images and
I think the test is very interesting. The comparasion is clear and the
reason why you prefer to use the ProPhotoRGB colorspace is obvious.

I am in the panorama business, for ech panorama I stitch a
equirectangular image from several images (input and output all
TIFF). In practice there is always some post processing neccesary
(althought now less then in the past thanks to a fine camera profile
and better settings control in RAW Converter 4.1) so working with a
larger color space then sRGB can indeed give a better result (but not
always). There is one downside when working with 16 bit source images;
the stitch process takes a much longer time then with 8 bit images.
The choice for 16 <> 8 bit depends on how much post processing is
needed to justify the extra time of 16 bit processing.
Did you test ever tested the differences between 16 <> 8 bit as well ?

Regards, Wim.

[Steve Sprengel]

With 8-bit there is more posterization in the shadows, but I wouldn't use 8-bit with ProPhotoRGB because any wider colorspace gets the bits it uses at the high and low ends of the histogram from the midtones, and with 8-bits there aren't that many colors to begin with, so you might actually see some banding in the midtones.  You can think of colorspaces as being like big bags of colornumbers and the number if bits is the density of colornumbers inside.  ProPhotoRGB is roughly twice as big as sRGB volume-wise so if you take 256 colors/channel in 8-bit sRGB and put them into a larger color-number bag, then the overall density will be less, about half as much.  With 16-bits there are 65536 colors/channel so changing the halving the density does not harm anything when the original images only had 4096 colors per channel with the 12-bit RAW and there are 16-times the colors in 16-bit as the original 12-bit RAW had.

 

I do some panoramas, just as a hobby not a business, and I don't use 16-bit ProPhotoRGB with them.  The images I start with could be manipulated with such a colorspace--Lightroom already uses a wide colorspace internally, but I export them as 8-bit sRGB to input into the panorama software. 

 

There are a couple reasons for doing this:  speed and size-of-temp storage is one reason, but also, I most do 360s with an 18mm lens, so it takes 60 images or so to fill out the entire sphere.  In any one of those images the dynamic range is not so much, so I can get away with only 8-bits for the segment of the panorama that image is for.  When the panorama software corrects the color and exposure differences it can end up with a wider colorspace overall and I may very well export the panorama to 16-bit TIF (or PSD) for final manipulation in Photoshop, but when I save for QuickTime, I again make it 8-bits. 

 

Some of my 360s are here:  http://www.sprengels.com/qtvr/

 

 

----- Original Message -----

From: MacUser

To: AcrCalibrator

Sent: Saturday, June 23, 2007 3:52 AM

Subject: [acrcal] Re: best color space for AcrCalibrator

[MacUser]

Hello Steve,

Nice to read that you ar familiar with making panorama's.
I looked at some of yours, you shot them with 18mm and needed a lot of
images.
The result is an high res pano with a lot of pixels so I guess you
downsized the stitched image for the web.
Although they contain lots of little stitch errors the results are
fine.
I am sure you know what caused the errors and if not I will explain.

Your explanation of the colorspace and the bit part is clear and it
explains why you (and I) can get away with 8 bit images.
I make panos intended for web use with much less images, between 3 and
10 with a shaved (removed sun cap) Tokina 10-17mm fish zoom lens on a
EOS5D.
Each image contains a big part of the panorama and therefore the
dynamic range in each image can be huge, I only can get away with sRGB/
8bit if the preprocessing of the RAWs is done as good as possible. By
doing this I only need little postprocessing of the pano. The only
kind of panos that need a lot of postprocessing, and therefore a large
colorpspace+bit depth, are HDR panos but I don't make them often
because they are very time consuming to make and the balance "time
+efforts <> quality gain" is often not right.

Since you are better colorspace experts of the two us I like to know
if you explain this:
Recently I compared the stitched output of 3 panos. 1 was made of a
set of sRGB/8bit images, 1 of AdobeRGB/8bit and 1 of AdobeRGB/16bit.
All source images+ stitched output are saved as TIF.
The shots are RAW with the same aperture+shutterspeed and all shots
are processed with same ACR parameters so all sets look the same.
After stitching and converting the 2 AdobeRGB's panos to sRGB I
compared the 3 panos.
The only difference I could see was a slight colorshift in the darks
of the pano made with the sRGB shots.
I guess this is because some colors are clipped during the stitch/
blend process because they got out out of sRGB space but the stitch
application (PTGui) don't handle colorspaces nor tags them so I wonder
how clipping can occurs. Any ideas ?

Best Wim

[MacUser]

PS,
here you can find examples of my work
http://www.dmmdh.nl

[Steve Sprengel]

With my 18mm I typically have 5 rows of vertical shots every 30 degrees for the middle three rows and every 40 degrees for the highest and lowest rows, and then a few shots at the bottom after I've removed the tripod.  I usually have around 60 pictures to stitch which is beyond the capabilities of my 5-year-old machine, sometimes.  I do resize the rectilinear image down to 8000x4000 before creating the QTVR.  I will look at your panos when I get home from work.  Viewing full-screen images across a remote-desktop connection is too slow to be enjoyed.

 

I use a cheap Panosaurus pano head so parallax shouldn't be too much of an issue, but the automatic focal-length and lens-distortion calculations of AutoPano can introduce some errors which are hopefully mostly mitigated by SmartBlend.  The latest beta of AutoPano purportedly has some improvements in lens-distortion calculations that I will be testing in the next few days.

 

A wider angle lens would help immensely and there are several rectilinear choices in the 10-20 range on my wish list but right now my main "zoom" is in for a repair estimate, and depending on how much they want, I may need to replace it, first.  I am also saving up for the new 40D that should be on sale by Christmas once it's announced this fall.

 

As far as the shadow areas looking different between an AdobeRGB and an sRGB panorama, I would guess it's the clipping in the smaller sRGB that's making a difference, but perhaps PTGUI amplifies this difference when it blends or corrects for color differences?  If PTGUI doesn't do much in the way of color-correction, then I would expect you could see similar differences in the input images between the sRGB and the AdobeRGB sets.

 

AutoPano is also adding support for fisheye and enhanced layered panoramas where bracketed exposures can be layered, and the whole thing is stitched at once, but you can output the LDR layers separately to be HDR-combined in Photomatix or whatever other tool.  AutoPano has been doing HDR internally for several versions, but sometimes Photomatix does a better combining job with the LDR layers.  With 18mm I typically shoot fixed aperture auto-exposure and just let AutoPano blend things but with a wider angle lens HDR would be more necessary.

 

http://en.wiki.autopano.net/Version_history

 

This thread doesn't have much to do with ACRCalibrator anymore, except I have created calibrations I use during my RAW development, so if you want to have this conversation offline, just reply to the e-mail address this comes from.

 

----- Original Message -----

From: MacUser

To: AcrCalibrator

Sent: Tuesday, June 26, 2007 2:44 AM

Subject: [acrcal] Re: best color space for AcrCalibrator

[Steve Sprengel]

I finally remembered to look at your panoramas when I was at home, and while I like the outdoors places as places to visit, I think I like the indoor and city panoramas the best, due to the manmade design vs the randomness of the outdoors textures.  I never have cared for the nadir logo covering the tripod area, though.

 

----- Original Message -----

From: MacUser

To: AcrCalibrator

Sent: Tuesday, June 26, 2007 2:46 AM

Subject: [acrcal] Re: best color space for AcrCalibrator

[MacUser]

Hello Steve,

Some like indoor panoramas while other prefer outdoor scenes, its a
matter of taste.
The logo is for promotional purposes and for covering the footprint of
the tripod and the panoramahead, a solution that can saves a lot of
postprocessing time and the need of an extra nadirshot.
BTW, the 40D will bring your panorama resolution to a very high level,
perhaps way to high so if possible do yourself a favour and buy a wide
angle fish lens, with the 40D+Tokina 10-17mm (standard, unshaved) you
can make panos of ca. 12kx6k (10mm) and even much higher when zoomed
in.

Thanks to your comment about the color shift of the sRGB
equirectangular I found the reason for the color shift.
I found out that the slight color shift of the sRGB pano all had to do
with the conversion of the colorspaces from the other panos with
adobeRGB and ProPhotoRGBto to sRGB. Both converted images look te same
so I thought that the unconverted sRGB pano was different. I guess
this can happens always if a colorspace conversion has to be made and
the fact that Adobe offers several methods for the conversion
(perceptual, colormatrix and others, with lots of options) makes it
obvious that there is no single "Use one, Fits all" colorspace
conversion method.

After trying some other options I find the perceptual method giving
the best results in this particular set of images, the colorshift of
the colorspace converted images was very small so in fact all
conversion methods are usefull.

Steve thanks again for you input, regards Wim.

[Steve Sprengel]

I do several 18mm nadir shots at different angles so SmartBlend can get rid of my feet.  I just exclude the tripod shots, altogether, from the render, after having used them to stitch the whole thing together.  I don't have much patience for cloning and masking and warping a nadir shot into place, either, but I usually get good enough results with SmartBlend.  A brick floor would probably require some post-processing.  Maybe the automatic stitching of nadir shots is more difficult if the lens is much wider angle.

 

I just got my estimate today on fixing my main walk-around lens and it's only going to cost $150 so I'll not have to replace it, and can decide on which wide-angle lens to get.

 

I think I'd much rather have a rectilinear lens than a fisheye.  It's less work in the software, at least, and the images look more "normal" if you're just taking wide-angle shots.  Is there something better about using a fisheye compared to a rectilinear of the same focal-length?

[MacUser]

Hello Steve,

> Is there something better about using a fisheye compared to a rectilinear of the same focal-length?

The answer is yes !

On my 350D I use a Canon EFS 10-22 lens, its is not suited for the 5D
but it is for your 300D and future 40D.
Its a little more expensive then a Sigma10-20 but defenitely worth the
extra money.
I have this lens for almost 2 year and it is my absolute favorit for
the 350D.
The lens is also usefull for making panoramas when the camera is in
portrait position with a minimum of 12 images @10mm (3 rows, 3+6+3)
although 16 and 18 images are much easier to stitch (4+8+4 or 6+6+6).

Happy shooting, Wim.

[Steve Sprengel]

I am confused...

 

Three lens have been mentioned:

 

Tokina 10-17 Fisheye (full-frame) $560

http://www.bhphotovideo.com/c/product/468737-REG/Tokina_ATX107PRODXC_10_17mm_f_3_5_4_5_AT_X_107.html

 

Sigma 10-20 Rectilinear (APS-C) $499

http://www.bhphotovideo.com/c/product/381610-REG/Sigma_201101_10_20mm_f_4_5_6_EX_DC.html

 

Canon 10-22 Rectilinear (APS-C) $689

http://www.bhphotovideo.com/c/product/351542-USA/Canon_9518A002_EF_S_10_22mm_f_3_5_4_5_USM.html

 

There is also a Tamron (APS-C) $569 lens of similar focal-length range:

http://www.bhphotovideo.com/c/product/363807-REG/Tamron_AF013C700_11_18mm_f_4_5_5_6_Di_II_Lens.html

 

In the previous e-mail you said good things about the Tokina fisheye.

 

So I asked if fisheye lenses are better than rectilinear lenses in terms of creating panoramas.

 

You said "Yes" but then went on to say a particular rectilinear lens was better than another rectilinear lens.

 

Could you explain about a fisheye or rectilinear being better for panoramas or was that not what you answered "Yes" to.

 

BTW, I am currently trying to stitch a 176 image 3-bracketed-exposures-per-position full-spherical panorama for HDR toning that seems to be a little too big for my computer to handle and of course the fewer images from an ultra-wide-angle would be a much easier to deal with.

 

 

----- Original Message -----

From: MacUser

To: AcrCalibrator

Sent: Saturday, June 30, 2007 3:03 AM

Subject: [acrcal] Re: best color space for AcrCalibrator

[MacUser]

Hello Steve,

I think I was not clear enough so I will explain my advice to you,

For making panoramas I prefer the Tokina10-17 lens because it provides
a much larger angle then a normal lens, so less images are required
and that is a big advantage.
I have experineces with 2 other fisheye lenses (Sigma+Nikon) and the
Tokina is the best.

For making panorama's AND for making normal rectangular images my
advice is the Canon 10-22.
A fisheye lens (zoom or not) is not suited for normal use, every image
has to be converted to a rectangular format and that is defenitely an
annoying job and therefore you must forget a fisheye lens if you want
to make normal images as well. Fortunately a normal wide angle lens is
also very suited for creating panoramas so you only have to make some
more images compared with a fisheye lens, thats all.
The Canon 10-22 has an extremely wide angle and @10mm it is very
usefull for creating panos with as little images as possible.

I only mentioned the Sigma10-20 because a lot of people buy this
popular lens because it has (almost) the same specs and is much
cheaper then the Canon 10-22.
If you can spent the money go for the best. I don't have any
experience with the Tamron lens, but I am sure there will be reviews
online.

Steve, your plan for stitching a panorama from 176 images (even if
they are shot bracketed) is defenitely not my cup of tea.
Unless the panorama is intended for printing I try to keep the number
of images as small as possible.
For web display I never need more then 10000x5000 px and normally I
stick to 7000x3500 px.
Even for printing 176 are a lot of images and I am sure you will get
an huge sized equirectangular.
I am very curious for what purpose you need such a panorama for.

Best Wim.

On Jun 30, 11:23 pm, "Steve Sprengel" <s...@sprengels.com> wrote:
> I am confused...
>
> Three lens have been mentioned:
>

> Tokina 10-17 Fisheye (full-frame) $560http://www.bhphotovideo.com/c/product/468737-REG/Tokina_ATX107PRODXC_...
>
> Sigma 10-20 Rectilinear (APS-C) $499http://www.bhphotovideo.com/c/product/381610-REG/Sigma_201101_10_20mm...
>
> Canon 10-22 Rectilinear (APS-C) $689http://www.bhphotovideo.com/c/product/351542-USA/Canon_9518A002_EF_S_...
>
> There is also a Tamron (APS-C) $569 lens of similar focal-length range:http://www.bhphotovideo.com/c/product/363807-REG/Tamron_AF013C700_11_...

[Steve Sprengel]

I entertain myself creating full-spherical 360s, mostly, and the 176-image panorama is one of those--to experiment with HDR.  There is no printing, only QTVR.  The 176 comes from using 18mm and 3-shots-per-position.  It's not divisible by three because I ran out of CF space 2 shots into the 3 of the last nadir shot.  Obviously if I had a ultra-wide-angle I'd only need a few dozen shots for the same coverage.

 

I was not considering a fisheye because of the look of the distortion.  The newest AutoPano-alpha has a fisheye mode where you define the image circle and it converts the images before they are stitched.  I am guessing the image data is not cropped to a rectangle and therefore would provide the most overlap, but perhaps I am wrong, not having tried it.

 

Is the difference between the Canon and the Sigma sharpness, or edge-aberrations (chromatic, coma, etc) or what is better in your experience?  I am leaning towards a cheaper UWA solution so my budget will have the capacity for the 40D, sooner.

 

 

----- Original Message -----

From: MacUser

To: AcrCalibrator

Sent: Monday, July 02, 2007 6:58 AM

Subject: [acrcal] Re: best color space for AcrCalibrator

Hello Steve,

[MacUser]

Hello Steve,

I don't have any experience with AutoPano.
For stitching panoramas I use PTGui because it is pretty advanced,
fisheye lenses, crop circles etc. etc. are supported and the Pro
version handles HDR very good.
The output can be set to several HDR and LDR formats, layered and
blended, whatever you prefer.
There is fine and active user forum and the developer provides good
help.

I prefer to use PTGui 7 Pro together with the latest beta(2) of
PhotoMatix 2.5 Pro for HDR panoramas, this combo is very good although
the HDR tone compressor of the latest PTGui beta is also very good.
At this moment version 7 of PTGui is available as beta and very soon a
final version will be released. I can advice you to download a free
copy of the application (http://www.ptgui.com) and give it a try.

I had short experiences with the Sigma10-20, after I bought it I was
not satisfied about shapness and the CA and therefore I changed it for
a Canon 10-22.
To be true the diffences between both lenses are small and perhaps if
there wasn't a cash-back action for the Canon lens at that time I
would kept the Sigma lens.
I am sure that there must be good comparing tests online because both
lenses are popular. Success with making your choice.

Best Wim.