Photo restoration glossary
The terms that come up when you start fixing photographs, in plain language, with a link to the page that covers each one properly.
- ArtefactDetail the model invented that was not in the original.
Any mark introduced by processing rather than recovered from the photograph. Common ones are matched glassy eyes on restored portraits, smeared patterns on fabric, and a waxy texture where skin has been smoothed past the point of realism. Artefacts hide well at print size and show under a magnifier.
How AI restoration works →- ColorizationAdding plausible colour to a monochrome photograph.
A model identifies regions such as skin, foliage, sky and fabric, then assigns each the colour it has most often seen for that context. Skin and landscape are reliable. Clothing, uniforms and eye colour are frequently invented, because a green dress and a navy one render as the same grey.
Colorize a black and white photo →- Compression blurSoftness caused by a file being re-saved, not by the camera.
Encoders discard detail to make files smaller, and each re-save discards more. A photo forwarded through four chats has been through four rounds of it. Unlike camera blur, the information is gone rather than displaced, which is why sharpening fails and upscaling is the tool that helps.
Fix a compressed photo →- DeblurRebuilding edges lost to movement or a missed focus point.
The model reconstructs the edges a photograph was meant to have. It works because blur from shake or focus error spreads real detail across neighbouring pixels rather than deleting it, so there is something to gather back.
Unblur a photo →- DPI and PPIPixels per printed inch, which decides how large you can print.
Divide the long edge of your file in pixels by the print size in inches. Around 250 is comfortable in the hand, and 150 to 200 is fine for a framed photograph viewed from across a room. The 300 figure people quote comes from magazine printing.
Printing and enlarging →- EmulsionThe light-sensitive layer that carries the image on a print.
Photographic paper holds its picture in a thin coating on the surface. When a print sticks to glass or to an album page, pulling it free lifts that layer and destroys the image permanently, which is why a stuck print is a job for a conservator rather than a kitchen table.
Handling damaged prints →- Face restorationA model trained specifically on portraits rather than photographs generally.
Human vision is unusually sensitive to faces, so a general sharpening pass that suits a landscape reads as unsettling on a person. A face model knows an eye contains a pupil, an iris and a catchlight, and rebuilds accordingly.
Restore faces →- FadingLoss of contrast as photographic dyes break down with light and age.
Dye layers fail at different rates, so colour prints drift as well as flatten. Cyan usually goes first, which is why so many 1970s photographs turn orange. Because the eye reads edge contrast as sharpness, a faded print also looks blurrier than it technically is.
Restoring faded photos →- Motion blurA directional smear from the camera or the subject moving.
The shutter stayed open while something moved. It smears in one direction, often with a faint double edge, and is one of the more recoverable kinds of blur because the detail is present and merely displaced.
Make a blurred picture clear →- PixelationVisible blocks where a file was saved with too few pixels.
The encoder averaged whole regions into single squares. Sharpening finds the boundaries between those squares and hardens them, producing something like stained glass, so upscaling is the correct response.
Unpixelate an image →- Remaster pictureSamsung Gallery's built-in one-tap enhancement.
Reached from the photo viewer menu or the edit screen, depending on your One UI version. It sharpens, lifts contrast and reduces noise in a single pass, and it handles recent phone photographs well. It is not tuned for old scans or physical damage.
Unblur on Android →- RestorationRepairing age damage, as distinct from removing blur.
Covers fading, cracks, scratches, stains and low resolution together. An old print usually presents several of these at once, which is why running a deblur pass alone on a 1962 photograph disappoints: it addresses one problem out of four.
Restore old photos →- SharpeningRaising contrast at edges that already exist in the file.
A conventional filter, not an AI operation. It cannot create an edge that is missing, so applied to a genuinely blurred photo it hardens the smudge and amplifies noise. This is the difference people run into when they expect a slider to unblur something.
- UpscalingGenerating new pixels to increase resolution, rather than stretching existing ones.
A resize spreads the pixels you have over a larger area, which is why enlargements look soft. An upscaler generates detail consistent with what it has learned about the world. Up to about 4x holds together well; beyond that faces are the first thing to look synthetic.
Upscale an image →