crud... I came up with (and executed) that idea four years ago, but it worked using a series of perl scripts piped together, no way I could deploy that to high school chem labs. By the time I learned how to program android, there were other, more capitalizable ideas...

The way it worked was to instead to use a 96-well plate (indexed, by blotting out certain wells) that would allow the software to key the matrix of wells and determine 'circles' to sample optical density based on the rgb values of the resulting jpg image - averaging across the entire circle. The experimenter would set up a 'calibration curve' in certain pre-indiced wells.

Typically you would have to have a color filter - i.e. if the color being evolved was yellow then you would take the photograph in blue light.