Restoring an old class or group photo
A group photo splits a limited pixel budget across every face — why the back row blurs, why a 600 DPI scan helps, and when to restore by row instead.
A class or group photo asks something a single portrait never has to: the same restoration has to work for every face in the frame at once, and those faces are rarely the same size. A face in the front row might span a couple hundred pixels; by the back row of a forty-person class photo, the same face can shrink to a few dozen, simply because it sits farther from the camera. This guide is about restoring the whole photograph honestly — what the scan needs to give it a real chance, why splitting the group into sections helps, and where a face is just too small to bring back.
Why the back row goes soft
Restoration works from the pixels a face actually has. A face spanning a couple hundred pixels gives real structure to work with — an eye's shape, a jaw's angle, skin texture. One spanning twenty or thirty pixels gives almost nothing: a smear of tone roughly where a face should be. That makes restoring a small face a different task from restoring a soft or damaged one at normal size — the print never recorded much detail there to begin with, so there is nothing hiding under blur or scratches to reveal.
Group and class photos concentrate this problem, because fitting everyone into one frame gives every face a smaller share of the picture than a single portrait would. Add a few rows of depth and the back row, already getting the smallest share of the frame, is also farthest from the camera — losing ground twice over. That is the real reason old class photos so often look sharp up front and indistinct at the back.
Worth telling apart from ordinary blur, its own topic in the guide to fixing a blurry old photo: a softly focused or shaky photo still recorded real information. A face small because the group barely fit in the frame never had much pixel data to begin with.
What a better scan buys you
The most direct lever comes before restoration enters the picture at all: how the print gets digitized. Our guide to digitizing old photos recommends 600 DPI for prints up to postcard size and 300 DPI as the floor for anything larger — a sensible baseline for a single portrait, where the whole print is essentially one subject. A group photo changes that math, because whatever resolution gets captured then has to be divided among however many faces are in it. Scanning a large group photo at 600 DPI or higher, rather than settling for the 300 DPI floor, directly raises how many real pixels land on each face once the print is that crowded.
Under-scanning does not cost a group photo evenly. A slightly soft scan of a solo portrait costs the whole photo a little sharpness; the same softness on a forty-person photo costs the back two rows disproportionately, since they already had the fewest pixels in the frame. Where a scanner offers a choice, the higher setting is worth the extra time — it adds real information rather than just processing what exists.
Restoring in sections, and the honest ceiling
Scan resolution is only half the story, because the tool's own output has a limit: a restored file comes back at roughly a thousand pixels on the long side, whatever went in. Run a whole forty-person photo through in one pass, and that budget spreads across the entire print — each face keeps a similarly small share of the output it had in the scan. A better scan gives restoration good material to work from, but does not change how the output frame is divided once forty people are sharing it.
The workaround is to stop treating the group as one image. Cropping a single row, or a small cluster of people, out of the original scan and restoring that section on its own puts the same roughly-thousand-pixel output on six or eight faces instead of forty — each comes back proportionally larger, no longer competing with everyone else for the same finite output.
This has a real ceiling. Restoring by section makes better use of the pixels that exist — it does not create pixels that were never captured. A face already only fifteen or twenty pixels across in the scan does not hold enough structure for any crop to recover; at that size, a confident result is an invented face, not a restoration, and an honest process declines to hand back a guess dressed up as one.
It is also a different job from pulling one person out of a group photo for a funeral program, covered in the guide to memorial photos, which crops and restores a single face. This guide restores the assembly itself, every row, without narrowing to one person first.
Restoring a class or group photo, row by row
Scan the whole print at the highest resolution you can
Use a flatbed scanner set to at least 600 DPI, even where the print is larger than postcard size and the general digitizing floor would allow 300. A crowded photo has less pixel budget per face, so the extra resolution goes further here.
Look row by row before restoring anything
Zoom into the scan and check each row. Note which faces are already a reasonable size and which are small enough to worry about — usually the back row or the edges.
Restore the whole photo first
Run the full scan through restoration as a baseline. For a smaller group, this alone is often enough, and it gives a repaired scene to compare everything else against.
Restore the smallest faces as their own crop
For rows that came back soft, crop that section from the original scan — not the already-restored whole photo — and run it separately for the most real detail to draw on.
Keep the results as separate files, not one merged print
The whole-photo restoration and the row crops stay separate images, not one stitched file — the whole photo for a group print, the row crops for a closer look at one part of it.
Group and class photo questions
How do I restore an old class photo with a lot of people in it?
Scan at the highest resolution your scanner offers, restore the whole print first, then check which rows came back soft. Crop those sections from the original scan and restore them separately — a smaller crop gives each face a bigger share of the output than the full group could.
Why do the faces in the back row look worse after restoration than the front row?
Because they had less to work with before restoration started. Faces farther from the camera occupy fewer pixels in the print, so restoring the same photo again does not add detail that was never captured. A higher-resolution scan and a separate crop for the back row both help, but neither fully closes the gap if the original faces were very small.
Can AI restore a group photo with many faces at once?
Yes, and for a modest number of people at a reasonable size, one pass is often enough. For a large class photo with faces down to a few dozen pixels, the output's own size is divided across everyone in the shot, so restoring rows separately usually helps.
Is there a limit to how small a face can be and still get restored?
Yes. A face occupying only fifteen or twenty pixels in the scan does not hold enough real structure to recover — what comes back at that size is closer to an invented face than a repaired one. Scanning higher before that point, and restoring smaller crops, keep a face on the right side of that line.
See how each row comes back before you decide
Upload the whole scan, or a cropped row, and look at the restored faces next to the original. In early access restoration is free, and nothing is charged until you are satisfied.
Epoha is in early access: restoration is free and there is nothing to pay yet.
Sources
- Preservation Guidelines for Digitizing Library Materials — Library of Congress