11 Replies Latest reply on Aug 4, 2008 11:15 AM by MadManChan2000

    Matrix and Lookup Table Profiles with the PE

    Bill_Janes Level 2
      This paper by Kopie concerns Foveon sensors, but the same principles apply to cameras with Bayer arrays. A matrix conversion is usually done to transform the camera color space to a working space such as XYZ. The matrix coefficients depend on the spectral qualities of the RGB filters of the Bayer array. The filters may not have a spectral response similar to the response of the RGB sensors in the eye and this can lead to problems. However, if the filters have a linear response, it is possible to produce an exact metameric match with the matrix conversion (previously stated by Thomas Knoll on the this forum).

      http://billjanes1.home.comcast.net/~billjanes1/binary/KopieSensorChar.pdf

      In the real world no such linear filters exist and the coefficients are usually chosen to produce the best overall results. Some colors may match and others will be off, and it is possible to calculate a metamerism index as shown in the paper. It would be easy to derive such an index from results with a Macbeth ColorChecker. If you used a ColorChecker DC, which has more colors, the process would be more involved. As you mention, results with real world images in the field are not readily available.

      As I understand the calibration process using the Fors script and similar methods, one is merely tweaking the coefficients used for the matrix conversion. Still some colors will be off, and one may gain additional accuracy by using a lookup table and this process is used by the Adobe ProfileEditor.

      Many camera profiling packages use the ColorChekerDC, which has many more colors, and I would expect this would give better results. I wonder if Adobe is going to go in this direction?