An input 2x2-pixel area at top left is converted to a 4x4-pixel RAO image bottom
left. Averaging each 2x2 of the RAO image yields the 2x2-pixel area immediately
below the input image. To the immediate right of those two images, the average
of the input image and the RAO image to a single pixel yields the RGB
percentages shown. The RAO Red is 99.4% accurate, the RAO Green is 99.6%
accurate, and the RAO Blue is 95.7% accurate to the input single pixel, for a
98.3% overall accuracy. I believe the largest possible inaccuracy in RAO using
this method is 95.7% per color component and composite. Zooming out to larger
square pixel areas increases accuracy.
