What an Image File Remembers About Its History
A PNG showing block artefacts was a JPEG at some point. That is a reliable reading whatever the extension now says.
An image file remembers what has been done to it. With JPEG the record is the damage itself. An image carries a record of what has been done to it. Compression damage accumulates and never reverses. Software records itself in the file. The relationship between the dimensions and the size says what quality setting somebody chose. Most of it is visible once you know where to look.
Almost everything written about image formats is aimed at people publishing them. This is for the other case, where a file has arrived and the question is what happened to it before it got here.
Why JPEG Compression Damage Accumulates
Lossy compression throws away detail it judges you will not miss. The judgement is made against whatever it is given. That is the part with consequences.
Save a photograph as a JPEG and some detail goes. Open that file, change something small and save again. The compression now runs against an image that already has damage in it. It compresses the damage as though it were detail and adds more of its own.
original fine detail intact
save 1 slight softening, invisible
save 3 faint blocks appear around edges
save 8 obvious blocking, colour fringing
save 20 visibly degraded, nothing recoverable
Nothing here is a fault in the format. JPEG was designed to discard detail and it does that well. The damage comes from asking it to do so repeatedly on its own output, which nobody intends and which happens whenever an image passes through several hands.
What JPEG Damage Looks Like
The artefacts are specific rather than general blurriness. Each one tells you something.
| What you see | What it means |
|---|---|
| Square blocks in flat areas | Heavy lossy compression, possibly several rounds |
| Haloes around sharp edges | A photograph format used on text or graphics |
| Colour smearing beyond an outline | Colour stored at lower detail than brightness |
| Banding in a smooth sky | Too few distinct values left to make a gradient |
| Soft edges on a large image | Enlarged from something smaller |
The third row is a design decision rather than a fault. Human eyes read brightness far more precisely than colour, so lossy formats keep less colour information on purpose. It works until an image has a sharp red line on white, at which point the colour visibly refuses to stay inside the shape.
The Conversion to PNG That Fixes Nothing
This is the commonest misunderstanding in the whole area and it is entirely reasonable.
What people expect
Lossless means gooddamaged JPEG
convert to PNG
PNG is lossless
therefore
the image is
now cleanReasonable. Also wrong. Lossless describes what happens next and says nothing about what happened before.
What happens
Faithful copyingdamaged JPEG
convert to PNG
every artefact
preserved exactly
result
same damage
larger filePNG records the pixels it was given. Those pixels include the blocks and the fringing.
A PNG showing block artefacts was a JPEG at some point. That is a reliable reading. PNG does not produce those patterns, so their presence means the image passed through a lossy format on its way here, whatever the extension now says.
What an Image File Says Beside the Picture
The image is one part of the file. Metadata sits alongside it and travels with it.
-
The cameraMake, model, lens, settingsWritten at capture. Shutter, aperture, whether the flash fired. Removed by many upload processes and kept by many others -
The softwareWhat has touched it sinceEditing programs leave a mark. So do phone processing pipelines and some messaging apps -
The datesTaken and modifiedTwo different things, often days or years apart. The gap is the interesting part -
The locationWhere it was takenPresent when the camera recorded it and the sender did not strip it. Precise enough to be a street address
Absent metadata is information too. A photograph with no camera details has been through something that removed them, which is most social platforms and some messaging apps. Untouched camera files are rich with this. A stripped file has a history even where the record of it is gone.
Read next A PDF Grows by Addition And Keeps What You Removed The same pattern in a document, where the history is even easier to read.Reading an Image File You Have Been Sent
- Check the extension against the file. The name is a claim, as with any other format. The first bytes say what it actually is.
- Compare the dimensions against the file size. A four thousand pixel photograph at two hundred kilobytes has been compressed hard. The same dimensions at eight megabytes has not.
- Zoom to the edges. Blocks and haloes cluster where detail changes sharply. Flat areas hide damage and edges display it.
- Read the metadata. Camera, software, dates. What is missing is as informative as what is there.
- Look for a lossless file with lossy damage. A PNG with blocking has a JPEG in its past. That usually matters more than anything else you find.
None of this is forensics. These are the things anybody can see in a couple of minutes with software they already have. Establishing what happened to an image in any formal sense is specialist work. It starts from the file rather than from the picture.
Keeping an Image File Properly
All of this collapses into a single habit. It is the same one that fixes several other problems on this site.
Keep the original untouched. Export from it every time rather than from your previous export. Then every version you produce sits one generation from the source instead of one generation from whatever came before.
The pattern is the same as the archive question and the conversion question. Work on a copy and keep the source. The worst outcome then is an afternoon spent rather than something lost. It is the closest thing to a general rule this whole subject has.
Why This Reads as Surprising
Images feel like the simplest kind of file. A picture is a picture. Either you can see it or you cannot.
What makes them awkward is that the damage is real, cumulative and mostly invisible until it is not. There is no error, no warning and no moment where anything goes wrong.
Nothing in the chain behaves badly. Somebody crops a photograph and saves it. Somebody else adds an arrow and saves it. A third person pastes it into a document that re-encodes it. Each step is reasonable and each one costs a little. By the time it matters the good version is several people back.
Behaviour checked against the referenced articles in August 2026. Reading an image for evidence in any formal sense is specialist work and nothing here substitutes for it.
Questions People Ask
7 questions, answered in full below.Does converting a JPEG to PNG improve it?
No. It stops further loss and undoes none of what already happened. Every artefact the JPEG had is copied faithfully into the PNG, usually into a much larger file, which is the worst of both outcomes.
Can I tell how many times an image was saved?
Roughly, by eye. The damage accumulates in the same places, so blocks around edges get harder and colour fringing gets wider with each round. It is an impression rather than a count. A strong one, once you know what to look at.
Why does my screenshot look smudgy?
Because it was saved as a JPEG somewhere. That format was built for photographs and handles sharp edges badly, so text and interface lines pick up haloes. Screenshots belong in a lossless format and often pass through something that decides otherwise.
What is in the metadata of a photo?
Typically the camera and lens, the exposure settings, the software that touched it and the dates. Sometimes a location. The picture shows none of it. It stays with the file unless something strips it out.
Does enlarging an image add detail?
It cannot. Detail that was not recorded is not in the file, so enlarging invents pixels between the ones you have. Modern tools invent convincingly. Convincing is not the same as recovered.
Can I tell if an image was edited?
Sometimes, though carefully. Editing software usually records itself in the metadata. Re-saving leaves compression damage in patterns. Neither is proof. Both are reasons to ask rather than conclusions to draw.
Which format should I keep an original in?
Whatever it arrived as, untouched, alongside anything you produce from it. That single habit prevents generation loss entirely, because every export starts from the same place rather than from the last export.
Sources
Where the figures and behaviour described above were checked.
- Generation loss Wikipedia
- Compression artifact Wikipedia
- Exif Wikipedia