One way you could do it is run your image into a top to chop which pulls one horizontal line of pixels and gives you a sample for each one. The crop tab lets you say what line of pixels you are pulling.
So you could lay down a top to chop for each line of pixels and use the samples it gives you to instance geometry and assign color.
You just need to figure out a way to assign each sample a x y position. Could be made with constant chops and beat chops or maybe written as an expression.
I think i am going to directly code it in openGL, using the CPlusPlus TOP.
but is there a simple way to use a SOP viewer, or something similar, to send its view matrix to my openGL code ? The idea is to be able to zoom, move and rotate the view inside the CPlusPlus TOP.