New project: camera color profile calculation - page 44 - RAW Shooting and Processing - CHDK Forum supplierdeeply

New project: camera color profile calculation

  • 480 Replies
  • 278126 Views
*

Offline vit40

  • ****
  • 276
Re: New project: camera color profile calculation
« Reply #430 on: 01 / June / 2009, 02:34:29 »
Advertisements
Hm, hard to say what is wrong. Can you post that dng (upload to www.megaupload.com or similar site), so I can take a look?
I planned to take some photos during this weekend with girl's camera to check the matrix and profile, but weather was bad, so we didn't go anywhere

Re: New project: camera color profile calculation
« Reply #431 on: 01 / June / 2009, 07:06:25 »
I took a .DNG picture directly from the camera and just used the SX110_ACR_4.4.cmd batch from vit40 on the dng file. Comparing the colors with the original jpeg file, they are almost the same, so maybe this is the right way? :)
Including the original jpeg and converted dng.


Original jpeg

From DNG

*

Offline chdkj

  • ***
  • 140
Re: New project: camera color profile calculation
« Reply #432 on: 01 / June / 2009, 14:31:03 »
Hm, hard to say what is wrong. Can you post that dng (upload to www.megaupload.com or similar site), so I can take a look?
I planned to take some photos during this weekend with girl's camera to check the matrix and profile, but weather was bad, so we didn't go anywhere
Uploaded 2 pics: a dng (http://www.megaupload.com/?d=B9W61A2I) and the corresponding jpg (http://www.megaupload.com/?d=KQAVUC1I). Both shows Sacre Coer in Paris. DNG is produced by DNG4PS from a RAW picture according to your instructions.
Thanks.

I took a .DNG picture directly from the camera and just used the SX110_ACR_4.4.cmd batch from vit40 on the dng file. Comparing the colors with the original jpeg file, they are almost the same, so maybe this is the right way? :)
Including the original jpeg and converted dng.
I think, it's the same color shifting. You took your pics during a cloudy sky, so the color shift might be not that distinct. It's much more heavier in the full sun.
Greetings.
« Last Edit: 01 / June / 2009, 14:33:42 by chdkj »
sx110

*

Offline vit40

  • ****
  • 276
Re: New project: camera color profile calculation
« Reply #433 on: 01 / June / 2009, 17:26:06 »
Thanks for the response

I took some research through the jpeg file and found white balance info in Canon maker notes, tag 13, although at different position than they were in jpeg from ewavr's camera (offset 80-83 in this case). WB multipliers are 889 1445 1945 889 (decimal), which can be translated to something like this:

dcraw -v -r 1.6254 1 2.1871 1  CRW_1356.DNG

What came out looks pretty much like jpeg picture, meaning that white balance of this color matrix is quite close to correct position (very slightly warmer, maybe 100-150 K - I didn't expect perfect match - for example, Coffin was more than 1000 K off with matrix for A620 ). Take a look (picture is quite noisy, far more than jpeg, as I didn't use any noise reduction)

http://www.megaupload.com/?d=1U1F4YFP

After short experimenting with ACR (because I don't know how to calculate color temperature from WB multipliers), it turned out that camera measured color temperature 4500-4600 K in this typical daylight scene (about 5200 K expected), that's why jpeg looks too "cold". On the other side, if you open this DNG with ACR and leave WB on as shot (which defaults to auto, because there is no WB info in this dng), it will measure 5850 K, quite different result, obviously because of different algoritm used, and if you leave it there, picture is very "warm". But setting temperature about 5000-5100 K gives, in my opionion, quite natural rendering of this daylight scene (because WB of matrix is moved slightly towards "warm"; if it was spot on, about 5200 should be about correct). If you want it to look "cold", similar to jpeg, set temperature about 4500 K, as camera measured in this case

Anyway, I noticed that sky hue is about 5 degrees too high (looks more violet that on jpeg, although for someone it will look more atractive), I moved it too far to the right from the position for sensor in Canon G9. Will correct this on next version of matrix, when I make some photos with the camera, when I get it. Foliage hue looks ok to me, slightly lower than on jpeg, like it is on Adobe color matrix for 400D, compared to jpeg. Red hue also seems to be ok, several degrees higher than on jpeg, as expected. In DNG profile I also included in previous post, sky hue is about correct

In short - don't worry, be happy ...


EDIT: new profiles in separate post --> http://chdk.setepontos.com/index.php/topic,3930.msg37314.html#new
« Last Edit: 28 / July / 2009, 16:21:05 by vit40 »


*

Offline chdkj

  • ***
  • 140
Re: New project: camera color profile calculation
« Reply #434 on: 02 / June / 2009, 01:32:03 »
He, He, He.... :-))

Hi vit40,

thank you for your efforts. I have to read your post again to make sure that I understand the whole thing ... :-). If I got you right than you could produce (working on) a color matrix that nearly fit perfect. If so then I can put those into camera color matrix and release the first stable version of CHDK for SX110 --> or you will do it, as you like.

The jpg-picture you linked above is really much more closer to the corresponding original jpeg. And the image section is larger than the jpg (you wrote it before but its funny anyhow).
I'm happy.

Greetings!
sx110

Re: New project: camera color profile calculation
« Reply #435 on: 02 / June / 2009, 13:27:41 »
OMG vit40 best ever!

SX110IS final is comming :)
Canon PowerShot SX110 IS
CHDK: at Autobuild now
Wiki: SX110IS

*

databoy

Re: New project: camera color profile calculation
« Reply #436 on: 02 / June / 2009, 21:48:05 »
The following free program may be useful.

http://www.nla.gov.au/preserve/dohm/driftt.html

*

Offline vit40

  • ****
  • 276
Re: New project: camera color profile calculation
« Reply #437 on: 03 / June / 2009, 05:01:16 »
Ok, guys

Yesterday, I tested WB of this matrix with WB info in 2 jpg images I made several days ago, one test image on CRT monitor, other through the windows (unfortunatelly, with not many colors), it seems that white balance of this matrix is quite good although not perfect and that WB multipliers are really at offset 80-83 of tag 13 of Maker Notes in jpeg. As I said, this is not final version, at least blue primary should be corrected, but it will be quite good for now I hope. You can change a matrix in old dng-s later, with DngSetMatrix

Anyway, for Photoshop users, I strongly suggest using DNG profile instead of embedded matrix (I will try to make another version in near future), because with embedded matrix, no matter how well calibrated it is (for instance - matrix for 400D, which is calibrated by Adobe, using professional equipments I suppose), some colors are quite different than they should be. In case of Adobe's calibration, red color is too orange (see red Skoda on Brasaral's jpeg) and sky color is usually considerably less or more saturated than on jpeg . With matrix, you can't get better result, because there are only 9 parameters to play with and only 6 of them affects colors, while 3 are for white balance. For instance, if you move red primary so that red is really red (hue = 0), than skin color isn't right, it's too red and doesn't look good. If you move blue pirmary so that blue is blue (hue = 240), than sky color if off ... etc. So Adobe introduced profiles ( = 4 matrices and HSV lookup table in an external file, with size 50-100 kB), and they explained why in their documents.


*

Offline zfeet

  • **
  • 67
Re: New project: camera color profile calculation
« Reply #438 on: 01 / July / 2009, 02:39:18 »
Here's the images for A430. At the moment when the pictures are processed with dcraw the results look purple. Also dng4ps2 crashes often when trying to process pictures taken with A430.

http://www.zshare.net/download/62076714495a8770/

*

Offline ArtDen

  • ***
  • 175
    • dng4ps2
Re: New project: camera color profile calculation
« Reply #439 on: 01 / July / 2009, 12:34:47 »
zfeet, image is too bright. Are you sure you set -2/3 for exposure compensation?

 

Related Topics