Color images are stored as three stacked grids rather than one. Each grid is called a channel, and the standard set is red, green, and blue. At every pixel location there are three intensity values, one per channel, each again ranging from 0 to 255 in 8-bit encoding. The displayed color is the result of combining those three values. A pixel with red 255, green 0, blue 0 shows pure red; red 0, green 0, blue 0 shows black; red 255, green 255, blue 255 shows white; and red 128, green 128, blue 128 shows a neutral gray. Because the channels are separate grids of the same size, a color image at 1920×1080 holds about six million numbers rather than two million. This matters for recognition: color is often a useful cue, but it is also the most fragile one, since the same object under different lighting produces very different channel values.
Learn MotionHow AI Sees: A Conceptual Guide to Image and Video Understanding
What a Machine Actually Sees
Three Numbers Make a Color
2 / 4
A color image is not one grid, it is three grids stacked on top of each other, one for red, one for green, one for blue. At any pixel position you read three numbers instead of one. Combine them and you get the color you see: full red with no green or blue gives pure red, all three at full value gives white, all three at zero gives black. Notice what this costs in data. A color image at nineteen twenty by ten eighty now holds about six million numbers. And notice the practical consequence: color is a strong cue, but change the lighting and those three values shift a lot for the same object.
0:00 / 0:00