Pixels, millimetres and the DPI that connects them
Almost every confusing resize problem comes down to one fact: a digital image has no physical size. It has a pixel count. A physical size only exists once you decide how many of those pixels occupy an inch, the resolution, or DPI.
That is why a form asking for a "3.5 × 4.5 cm photograph" is asking an incomplete question, and why the same requirement produces wildly different pixel counts:
| Resolution | 35 × 45 mm becomes | Typical use |
|---|---|---|
| 96 DPI | 132 × 170 px | Screen only, too small to print |
| 300 DPI | 413 × 531 px | The standard for photo requirements |
| 600 DPI | 827 × 1063 px | High-quality print |
If a requirement gives millimetres without a DPI, use 300. It is the near-universal assumption behind published photo specifications.
Why this tool writes DPI into the file
Browsers do not record resolution when they save an image, the JPEG density field is left as "aspect ratio only". So a 413 × 531 px photo produced by most online tools carries no evidence that it is meant to be 35 × 45 mm.
Usually that is harmless. It stops being harmless when the photo will be printed, or when a validator reads the metadata to check the physical size. When you resize in millimetres, centimetres or inches here, the resolution is written into the file so it reports its true physical size.
Choosing what happens when the shape changes
If your image is 4:3 and the target is square, something has to give. The three options handle it very differently:
- Crop to fill, takes the largest centred region matching the new shape. Nothing is distorted; the edges are lost. This is almost always the right choice for photos of people.
- Fit and pad, keeps the whole image and fills the remainder with background. Right for logos, documents and diagrams where nothing may be cut off.
- Stretch, squashes the image into the new shape. Never use this on a face; the distortion is obvious and fails photo review.
If you need to control exactly which part is kept, crop first with the crop tool, then resize.
Enlarging does not add detail
Resizing down discards information and looks fine. Resizing up has to invent information, and inventing it means guessing. A 200 px image scaled to 1000 px is a 200 px image with softer edges.
If a requirement demands more pixels than your source has, the real fix is a new photo, not an upscale. This is the most common reason a photo is refused for being "low quality" despite having the right dimensions.
The order that gives the best result
- Crop to the required shape, so you control what is kept.
- Resize to the required pixel dimensions.
- Compress to the required file size last.
Doing it in this order means compression is only ever working on pixels that matter. Compressing first and resizing after throws away detail twice and looks noticeably worse at the same file size. If your requirement is a standard one, the photo maker does all three in a single step.