Since the first PDS release came out, I've been struggling with the best way to display calibrated 3 filter color, especially those with the L2 as red. Most methods can create an internally consistent looking image from a particular 3 filter set, but don't quite match other filter sets of the same target. I recently started in on a new attempt, one which almost completely abandons all color theory and replaces it with a treatment the images strictly as numeric data. Taking the first 180 sols of 6 filter imagery as a training set, I use a least squares fit to come up with formula for each 3 filter combo which best reproduces the XYZ values calculated from the more accurate 6 filter fit. I knew up front this was a bit of a kludge, but was pleasantly surprised at the consistency it produced.
As an example of this new method, I've posted up an update to my site. The calibrated galleries now include ...all... the 3 filter (256,257,456,457,247) as well as 6 filter images from the first 180 sols, weighing in at a hefty 1598 color images from Opportunity, and 1741 from Spirit.
A more complete description of the methods used can be found here
Also, I feel it appropriate to mention here of all places, my images are very bland and aesthetically boring in comparison to the tremendous efforts put forth in this forum and by JPL. My goal has always been consistency and comparability. As we know from working with the stretched raw images, changes in the scale of brightness can cause major consistency problems when looking at two images side by side. With that in mind, no local or 'by image' contrast adjustments have been made, almost all images appear in the same scale of brightness.