On 2009 Sep 8, at 11:35 PM, Pascal de Bruijn wrote:
On Wed, Sep 9, 2009 at 12:28 AM, Ben Goren<ben@xxxxxxxxxxxxxxxx> wrote:On 2009 Sep 8, at 2:07 PM, Pascal de Bruijn wrote:I'm getting pretty decent results, using UFRaw and an IT8 target... soI'm guessing it should be possible for DCRaw as well...Pascal,Are you using linear RAW files, or are you getting ``close enough'' and thenrefining the results?No, I'm using linear RAW files... The results are pretty good, but not perfect... I'm using single-shaper matrix profiles, which have gotten me the best results thus far... My method: http://blog.pcode.nl/2008/11/15/color-profiling-your-own-dslr-redux/ I'm still refining it a bit... but the general idea works well.
I did a bit of experimentation and got some pretty good results:
profile check complete, peak err = 7.518615, avg err = 0.855233
I used UFRaw to get the channel multipliers for the white balance by selecting most of the area of one of the neutral patches.
Next, I used dcraw for the actual conversion:
dcraw -v -r 2.474 1.000 1.335 1.000 -o 0 -4 -T _MG_1397.CR2
I then created a profile with each type of algorithm, and discovered that an XYZ LUT profile gave significantly better results than any of the others -- according to the numbers, visually, and by checking with an eyedropper in Photoshop.
I'm sure I'll do at least some more experimentation, but this should be good enough. If anybody has any suggestions as to make it even better...I'm all ears, of course.
Thanks all for the help! Cheers, b&