HomeGuides › Why photos look blurry

Why Do Photos Look Blurry When You Enlarge Them?

The reason has nothing to do with your screen or your software. It is built into how digital images work.

Advertisement

An image is a fixed grid

Photographs feel continuous, like a window onto a scene. They are not. A digital image is a rectangular grid of coloured squares called pixels, and the number of them was decided at the moment the picture was taken or saved.

A 600 x 400 image contains 240,000 squares. That is the complete contents of the file. There is no hidden extra detail waiting to be revealed by zooming, in the way a film negative holds more than a small print of it does.

What happens when you drag it bigger

Display a 600 x 400 image at 1200 x 800 and the software has four times as many screen positions to fill as it has real pixels. It must invent three out of every four.

The traditional methods all work by looking at nearby known pixels and calculating something in between:

Every one of them is guessing based purely on arithmetic. None has any idea whether it is looking at an eyelash, a brick wall or a letter of text. Averaging is a safe guess, and safe guesses look soft. That is the blur you see.

Advertisement

Why compression makes it worse

Most photographs are JPGs, and JPG saves space by discarding detail. At low quality settings it leaves behind faint square blocks and coloured fringes around sharp edges.

At normal size these artefacts hide easily. Enlarge the picture and they are magnified along with everything else, so a picture that looked acceptable becomes visibly damaged. This is why a file that has been shared through several messaging apps enlarges so badly: each app re-compressed it.

How AI changes the guess

An AI upscaler replaces arithmetic with experience. The model has been shown millions of examples of small images alongside their sharp originals, so it has learned what real detail looks like at a larger size.

Given an edge, it does not average across it, because it has learned that edges stay sharp. Given skin, it reproduces skin texture. Given text, it reconstructs letter shapes. The output is still a guess, but it is an educated one, and it looks far closer to a real photograph than any averaging method can.

What still cannot be recovered

It is worth being clear about the limits, because no tool escapes them.

ProblemCan AI upscaling help?
Small image, otherwise sharpYes, this is exactly what it is for
Mild compression artefactsPartly, results vary
Out of focusVery little; sharpness was never captured
Motion blurVery little, same reason
Text too small to readNo, the letters are not in the file
Heavy noise or grainSometimes worsened, since noise gets sharpened too

The practical takeaway

Blurriness on enlargement is not a fault in your software. It is the honest consequence of asking for more detail than the file contains. The fix is either to start from a bigger original, or to use a tool that reconstructs detail rather than averaging it.

If you want to see the difference, upscale one of your own images and compare it against simply stretching the same file in an ordinary editor.

Advertisement