yes, you describe fractals correctly. but i fail to see them as a photo compression tool. if the camera only generated a single green pixel for an individual leaf, i fail to see how any algorithm can turn that into a leaf when you enlarge the picture. particularly when that same green pixel next to it is kermit the frogs head.
at the end of the day a camera pulls in a raster image (true for analog and digital – though analog is an insanely larger scale).