New project: camera color profile calculation - page 4 - RAW Shooting and Processing - CHDK Forum

New project: camera color profile calculation

  • 480 Replies
  • 319212 Views
Re: New project: camera color profile calculation
« Reply #30 on: 28 / January / 2008, 04:52:53 »
Advertisements
Well, to reiterate Rommel's question - where do we find the DNGSetMatrix?

*

Offline vit40

  • ****
  • 276
Re: New project: camera color profile calculation
« Reply #31 on: 28 / January / 2008, 05:08:40 »
Here's new link (the one in dpreview forum is dead, I think)

http://www.zshare.net/download/694882485879b8/

*

Offline ewavr

  • ****
  • 1057
  • A710IS

*

Offline ArtDen

  • ***
  • 175
    • dng4ps2
Re: New project: camera color profile calculation
« Reply #33 on: 28 / January / 2008, 06:11:07 »
OK, it seems that ArtDen changed the matrix in the meantime:
Yes, I recalculated matrix with another Jpeg gamma value. Red color looks better with this matrix.

Btw, do you normalize inversed matrix before testing into LAB area?
« Last Edit: 28 / January / 2008, 06:12:43 by ArtDen »


*

Offline vit40

  • ****
  • 276
Re: New project: camera color profile calculation
« Reply #34 on: 28 / January / 2008, 06:37:31 »
Well, for A620, I started with Coffin's matrix (just converted to xyY, so that data were meaningful for me), first corrected Y values for R and B to achieve correct WB, then moved position of primaries around (changed x or/and y; Y should be also corrected every time), until I got closest match on macbeth target, shoot with A620, to the result from 400D. Additionally, I corrected matrix on some real life photos, because ACR matrix for 400D isn't very good in red range - red color has too high hue, so everything is too orange (even on my CRT monitor, where pure red looks less orange than on LCD monitors) . I was checking photos in RGB and HSV, wasn't using Lab.

There's a part of code in DngSetMatrix with variable Normalized, but it's not normalization, it just keeps dividing all values with 10 until every value is <= 3, so that displayed values fit into output format with 4 decimal places

  repeat
    normalized := True;
    for i := 1 to 9 do
      if Matrix1 > 3 then normalized := False;
    if normalized then Break;
    for i := 1 to 9 do
      Matrix1 := Matrix1 / 10;
  until False;


Before writing to dng, every value is multiplied with 10000, so all (integer) values are in range -nnnn .. 30000. I suppose there is more elegant way to achieve this, but it works
« Last Edit: 28 / January / 2008, 07:24:50 by vit40 »

*

Offline ArtDen

  • ***
  • 175
    • dng4ps2
Re: New project: camera color profile calculation
« Reply #35 on: 28 / January / 2008, 11:43:35 »
Some changes in conditions. Before tacking shot exposure compensation have to be setted to -1/3 or -2/3.
It looks like too bright colors forces camera to apply stronger tone mapping for Jpeg image. Please take another photo with new condition and reupload it again. Sorry for unnecessary work.
« Last Edit: 28 / January / 2008, 11:58:11 by ArtDen »

*

Offline ewavr

  • ****
  • 1057
  • A710IS
Re: New project: camera color profile calculation
« Reply #36 on: 28 / January / 2008, 12:29:41 »

Re: New project: camera color profile calculation
« Reply #37 on: 28 / January / 2008, 12:44:11 »
rommel_a720is_2.zip First is with the original specs, then -1/3 and then -2/3


*

Offline jeff666

  • ****
  • 181
  • A720IS
Re: New project: camera color profile calculation
« Reply #38 on: 28 / January / 2008, 12:49:00 »
A720, -1/3

I also changed my monitor-settings a little so fields with different brightness are more distinguishable.

http://www.zshare.net/download/6960229ac700e0/

*

Offline ArtDen

  • ***
  • 175
    • dng4ps2
Re: New project: camera color profile calculation
« Reply #39 on: 28 / January / 2008, 13:36:47 »
A710 (-2/3 Ev) - http://www.zshare.net/download/69595732adc6cd/
Thanks! matrix for this image is
Code: [Select]
0.664401, -0.206085, -0.105935,
-0.208489, 0.655900, 0.051953,
-0.051591, 0.115760, 0.234946

 

Related Topics