The dynamic range of most display devices, and the data discrimination capabilities of most data formats, greatly exceeds the human eye's except in the range of identifying hues, where the eye reigns supreme.

We were doing very similar work in the late 70s with satellite scanner data. We used a Saturation-Hue-Intensity transform to create an image from the original RGB scene, applied a piecewise linear contrast stretch to the Hue band to enhance those parts of the histogram where the color was hiding, regular histogram equalizations to the Saturation and Intensity bands, and then transformed back into normal color space. We could transform dull, flat desert scenes from uniform browns and grays to wonderful subtle gradations , color textures and tones. The fact that these images revealed previously hidden drainage features and geological formations verified that it was not just artifacts of the processing.

Of course, the results looked nothing like what you would see if you just looked out the window at the ground. But the scene looked realistic and natural, it didn't look obviously processed like some color enhancements do. I we had more than three bands to begin with, we reduced them to three with a principal components analysis. But we always worked with just one image at a time. Multiple imagery taken at differnt times introduces all sorts of artifacts.

Although we had integer, floating point, and double precision (2, 4 and 8 byte) data available then, we could get very good results with this technique with just plain byte data (256 dN levels).

The best example of this technique you are probably familiar with were those original spacecraft images of Io, that looked a lot like a pizza. That same data, transformed in SHI space and processed there prior to being RGBtranned back, resulted in some beautiful and very revealing images of the moon, in beautiful shades of tan.