As I understand it (and I've never looked at one myself, just followed along in the blogs of some working astronomers), it basically has an EXIF the size of the original file. All sorts of data and history about each pixel of the detector. After all, an astronomer doesn't want to declare she's found a new supernovae because she got mislead by a cosmic ray hit on the detector.

How it is all listed and encoded I have no idea, but the working stiffs seem to describe it as looking more like a table (and that's the way they interact with it), than like a bitmap.

Gigabyte files are nothing new for the scientific community. They tend to store and send a LOT of data. And all forms of lossy compression are of course abhorred.

For all I know, though, some of the cameras on Curiosity are sending ordinary bitmapped images. And all you'd really need is knowledge of the response curves of the CCDs/the filters in use to reconstruct the original colorspace.

Always been a puzzle to me how exactly you'd go about displaying the "real" color of the Martian surface, though. I mean, to a person, not as a data representation. The "real" color where? If that dirt was Earth, under a blue sky? As seen under artificial light which wasn't in the original picture? To eyes that are adapted (as human eyes do) to the color correction of a studio environment? Or eyes that have color-adapted to the Martian environment?

Once you get outside of "54 lumens at 535 nanometers" and start asking what it "looks" like to those ever-changing and highly adaptable human eyes -- in lighting conditions that aren't anything like those found on Mars -- it gets very hard to define what is the "correct" color.