Morning Daniel,

I've noticed your work at lyle.org, and I think you're doing a great job.

It's true that we won't be able to get really accurate calibrations until the PDS dump. I'm looking forward to it--I've written a bunch of PDS code for MGS-MOC and a lot of it will be reusable for the MER images. With all the calibration data you just click a button and out pops a (relatively) perfect image. I'd discounted the Maestro data sets because looking at the first three releases it seemed like they included just a tiny fraction of the images. Now that you've pointed it out, I'll look there for what I can get to do a better job with at least some of the images.

In the meantime, while we wait for PDS, I'm just using an empirical technique to get into the ballpark. I assume that the cameras scale the data up to take full advantage of the dynamic range, based on the ground calibrated relative exposure times. Then I assume that any variation in overall brightness is a result of autoexposure, and back it out with a sort of "autounexposure". It seems to do a reasonable job judging by comparison with NASA's released images. According to Justin Maki, the label "approximate true color" signifies that they're doing only a rough calibration too.

Up until recently I was working in RGB, but that's soooo sloppy. This last weekend I started converting to CIE, and I think that will work much better, especially when creating true multispectral images from a set of more than 3 filters. So far I've rewritten my old 3-channel generator to use CIE, and I'm already seeing the effects of properly mapping wavelengths: L4-based images have more green, L2 images more red. CIE's the way to go.

Sounds like we're working roughly in parallel. I find that reassuring, to tell you the truth, since I'm making it all up as I go and learning a lot in the process.