dx, the original MER camera CCD data is actually 12 bits of data per pixel.
Twelve bits of data fits conveniently in 16 bits - ( Programmers, for reasons best not explained here, naturally think in powers of 2 ) - so that is why I work with 16 bits in imagej.
JPG RGB images are 24 bit data ( 8 bits per color channel). To see this, load any color RGB image into imagej and change the image type to RGB stack. Notice that the result is a stack of 3 8 bit images.
For reasons known only to the MER team, the MER public "raw" images ( RAW ) are usually JPG RGB B&W images.
The three color channels of the RAW data should be exactly the same - as they represent a B&W filter image - but sometimes they are not! And sometime they are 8 bit JPG images - as they all should be!!
OK, now to answer your question: No.
The reason is that if I have 16 bit image data I already have everything there is to know about the data.
The RAW data is massaged original data - and exactly what has been done to get from 12 bits to 8 bits of information is unknown ( to me ).
The best MER image data to work with - when it is available - is the RAD format data. It is intended for real scientific work and a number of transforms have been done on the data to normalize it for comparison with other images - and other data sets.
For my purposes, the removal of camera and CCD imperfections is the most useful transforms. The conversion of the data to spectral radiance units is almost useless for making "pretty pictures". I usually start my work with RAD images by converting the spectral data back to 16 bit exposure data!
Anyway, no. I haven't tried to combine the two.
Maybe there is a good reason ( like studying the compression and clipping used to create the RAW images and then trying to "back out" the original data from RAW data which is not yet available as original data. ) for doing so - but I haven't reached terminal boredom - yet.