Images as Numbers
Zoom into any photo far enough and the picture disappears into a grid of numbers with nothing else hiding inside it. Click through a tiny image pixel by pixel and watch a face turn back into the grid it always was.
Worth reading first: What Is Computer Vision?
You believe a photograph has a face in it somewhere, tucked inside the file, the way a letter has words tucked inside an envelope. It does not. Zoom in far enough on any photo and the face runs out — not blurs, not fades, runs out — and what is left is a grid of numbers with nothing else hiding behind it.
A grid with nowhere to hide
Open any image file and there is no layer underneath the pixels. No hidden field says “dog”, no outline marks where the ears are. A photograph is a rectangle of numbers, full stop. Whatever meaning you get out of it, you or a model built back up from that grid — the file never carried it.
The interactive below is a real image, not a metaphor for one: a 12-by-12 grid of 144 numbers that happens to render as a smiling face. Step through the four views and watch the same 144 numbers stop looking like a photo and start looking like exactly what they are.
The same 144 numbers, rendered four ways. Nothing is added or hidden between stages — only how closely you are looking changes.
Click any square above to read the exact number stored in that pixel.
What one pixel actually stores
Click any square in the grid above and the value it shows is the entire content of that pixel — one integer, from 0 to 255, for a grayscale image. Nothing else is stored there. 0 is black, 255 is white, and everything between is a shade of grey with no unit finer than a whole number: 255 possible steps and not one more.
A colour pixel is not one number, it is three — one each for red, green and blue, stacked on the same spot. That is enough of a different problem that it gets its own lesson next; for now, what matters is that whether it takes one number or three, a pixel never stores anything about what is near it. It has no idea it is part of an eye.
Resolution is a budget
Resolution is just a name for how many of those numbers you have. The smiley above is 144 numbers. A typical 4K photograph is 3840 pixels wide and 2160 tall — 8,294,400 pixels, each with three colour values, for just under 25 million numbers in one photo. A classic machine-learning thumbnail, the kind used for handwritten-digit recognition, is 28 pixels by 28 — 784 numbers.
Neither size is correct in some absolute sense. More pixels buys detail — the edge of a leaf, a strand of hair, a distant street sign made legible — and it costs storage, transfer time, and computation, because every one of those millions of numbers has to be read by whatever processes the image. A model built to run on a phone in real time is often deliberately fed a smaller grid than the camera captured, because the detail it would gain is not worth the numbers it would have to chew through to get it.
Zooming in until the picture breaks
Zoom into a real photograph on your phone and the same thing happens as with the grid above, just at a scale your eye is not used to noticing. Past a certain point, the picture stops looking like a face or a leaf or a street sign and starts looking like what the earlier “Pixel grid” and “Numbers” views showed you — flat squares of colour, each one a number, arranged in rows.
That is not the camera failing or the file corrupting. It is the picture running out — you have reached the resolution budget the photo was captured at, and there is nothing finer underneath to zoom into. Recognisability was never a property of the file. It was a property of the numbers being dense enough, at the size you were viewing them, for your visual system to stitch them back into a face.
Key takeaways
- A photograph has no hidden layer of meaning — it is a rectangle of numbers, and nothing else is stored in the file.
- One grayscale pixel is a single integer from 0 to 255. A colour pixel is three of them, stacked on the same spot.
- Resolution is simply a count of how many of those numbers you have, and it is a budget, not a virtue — more pixels costs more storage and more computation for every downstream step.
- A 4K photo (about 25 million numbers) and a 28-by-28 thumbnail (784 numbers) are the same kind of object at wildly different budgets.
- Zoom into any real photo far enough and recognisability collapses into flat squares of colour, because you have reached the resolution it was captured at.
Quick check
Answer these to unlock the next chapter — 3 of 4 to pass. You can retake it anytime.
Answer every question to check.
Make a free account to read on
Every chapter is free — an account is how your progress, XP, and streak follow you from your laptop to your phone, and how you show up on the leaderboard. No payment, no trial.