Nudging a rotation by a few degrees several times, resampling on each pass, which softens the image a little more every time.
Resolution and scaling
Interpolation
Also called resampling, image interpolation or resample. Here is what it means, when it changes what you export, and what people get wrong.
Resolution and scalingRuns on your deviceUpdated 10 August 2026
The short answer
Interpolation is how software works out pixel values at positions that did not exist before, which is what happens whenever you resize, rotate or warp an image. It estimates each new pixel from its neighbours using a chosen filter. Interpolation redistributes the information already present; it never adds more.
Updated
The short version
Interpolation at a glance
| Quick fact | Detail |
|---|---|
| Also called | Resampling |
| Happens during | Resizing, rotating, warping, lens correction |
| Common default | Bicubic |
| Best for shrinking | Lanczos or area averaging |
| Adds new detail | No |
| Exception | Learned super-resolution, a different process |
Slide the table sideways to see every column.
How does it actually work?
The program builds the output grid first, then asks where each output pixel would have sat in the input. Those coordinates almost never land on a whole number: an output pixel might map to input position 12.4, 7.8, which is between four known samples. The interpolation filter decides how to turn the surrounding known pixels into one value, and how many of them to consult. That single decision is the difference between a crisp resize and a mushy one.
Which filters are there and what do they trade?
| Method | Pixels read | Speed | Best for | Weakness |
|---|---|---|---|---|
| Nearest neighbour | 1 | Fastest | Pixel art, label maps, whole-number scaling | Blocky, aliases badly when shrinking |
| Bilinear | 4 (2 x 2) | Fast | Live previews, small adjustments | Noticeably soft |
| Bicubic | 16 (4 x 4) | Moderate | General photo resizing, the usual default | Slight ringing at hard edges |
| Lanczos | 36 or 64 | Slower | Shrinking photographs, holding fine detail | Visible haloes on high contrast |
| Area averaging | Every covered pixel | Fast | Large reductions | Not suitable for enlarging |
Why can interpolation never add detail?
Every output pixel is a weighted sum of input pixels. No matter how clever the weights, the result contains only combinations of measurements that were already in the file, so no new information enters. Enlarge by 4x and you hold sixteen times the pixels carrying exactly the same detail, spread thinner.
Learned super-resolution is a different process. It does add high-frequency content, drawn from patterns it saw during training rather than from your image. That is why it can look sharper than any interpolation filter, and also why it can be wrong in ways interpolation never is.
Which one should you pick?
Shrinking photographs: Lanczos or area averaging, because both read the whole region being collapsed and suppress moire. Enlarging photographs: bicubic for modest changes, a learned model beyond roughly 2x. Anything non-photographic with hard edges, such as pixel art, screenshots of text, QR codes or segmentation masks: nearest neighbour, since blending genuinely damages them. Rotating or straightening: bicubic, done once from the original rather than as a series of small nudges.
What people get wrong about interpolation
Each one is a real failure mode, not a style preference.
Using bicubic on pixel art or a screenshot of text, which turns crisp blocks into a grey smear.
Expecting a better filter to recover lost detail, when the filter only changes how existing information is spread.
Questions people ask
Interpolation, answered
The follow-up questions people search for once they have the definition.
What is the difference between interpolation and resampling?
In everyday use they are interchangeable. Strictly, resampling is the whole operation of rebuilding an image on a new pixel grid, and interpolation is the estimation step inside it that produces each new value. Editors label the setting either way.
Which interpolation is best for enlarging photos?
Bicubic among the classic filters, since it uses sixteen neighbours and keeps edges reasonably firm. Beyond about 2x, a learned super-resolution model will look considerably better, at the cost of inventing texture rather than estimating it.
Does interpolation happen when I rotate an image?
Yes, for any angle that is not a multiple of 90 degrees. The rotated grid does not line up with the original, so every pixel is re-estimated. Rotations of exactly 90, 180 or 270 degrees just move whole pixels and are lossless.
Why does my image get softer each time I resize it?
Because every pass averages neighbouring pixels, and averaging an average blurs further. Two resizes soften more than one, and small repeated adjustments are worst of all. Always resize once, from the original file, straight to the size you need.
Same cluster
More from the glossary
Neighbouring entries, so every term in the set is one hop from every other.
JFIF
JFIF (JPEG File Interchange Format) is the standard wrapper that makes a JPEG readable by other programs.
Inpainting
Inpainting fills a selected region of a photo with content that matches its surroundings, so whatever was there looks as though it never existed.
JPEG
JPEG is the standard lossy format for photographs.
From the studio
Tools that do this
Each one runs on your own device, so the photo stays where it is.
Keep reading
Elsewhere on Photomop
Platform image rules
Image size, file type and background rules for 22 selling platforms, each one sourced.
Read itConvert image
JPG, PNG, WebP and AVIF, with an honest answer about HEIC and alpha.
Read itAll the tools
The whole toolbox in three families, with what each one needs before it can run.
Read itOn-device privacy
The pipeline in five steps, what leaves your device, and how to check it yourself.
Read it
Now do it to a photo
Photomop is a photo studio that runs on your own device. Resize, crop, compress, convert, batch edit, remove a background and change one all work in the browser tab you are reading this in, at full resolution and with no watermark. The on-device page shows you how to check that the photo stays put.