Journal figure size and DPI checker and resizer

$$\begin{array}{rl}\text{pixels}& =\text{width (in)}\times \text{dpi}\\ & =\text{width (mm)}÷25.4\times \text{dpi}\end{array}$$$$\text{dpi at a width}=\text{pixels}÷\text{width (in)}$$$$\begin{array}{rl}& \text{printed width (mm)}\\ & =\text{pixels}÷\text{dpi}\times 25.4\end{array}$$

Pixels and dots per inch. A raster image has a fixed number of pixels; its DPI (more exactly PPI, pixels per inch) only says how many of them go into an inch when it is printed. One inch is 25.4 mm exactly, so a figure printed 89 mm wide is 3.504 inches wide, and at 300 dpi it needs $3.504\times 300=1051$ pixels across. The DPI number written in a file does not change the pixels: a 1051-pixel image saved at 72 dpi and at 300 dpi is the same image, but programs place the first one 371 mm wide and the second 89 mm wide. Journals judge the pixels at the printed size, which is why the same file can pass at a single column and fail at a double column.

How to do it by hand. Find the width in the journal’s guidelines, divide millimetres by 25.4 to get inches, and multiply by the dpi the journal asks for that kind of image. Divide your image’s width in pixels by the inches to get its resolution at that size. The pixels needed are rounded to the nearest whole pixel, as Elsevier’s own table does (90 mm at 300 dpi = 1062.99, listed as 1063).

Which dpi. Publishers set the resolution by the kind of image: photographs and other continuous-tone (halftone) images at 300 dpi; combination figures, where photographs carry text or line drawings, at 500–900 dpi; line art (graphs, drawings, black-and-white bitmaps) at 1000–1200 dpi, because thin lines and small text show their pixels first. Where a journal states no figure for line art, as Nature does, it asks for vector files (PDF, EPS or AI) instead, which have no resolution. The table below gives each journal’s own wording.

Guidelines used (checked 8 November 2026): Nature: final submission; Nature Research Figure Guide: building and exporting figure panels; Science: instructions for preparing a revised manuscript; Science: instructions for preparing an initial manuscript; Cell Press: figure guidelines; PNAS: digital art guidelines (PDF); PLOS ONE: figures; JACS author guidelines; Elsevier: artwork sizing; eLife author guide (PDF).

Resampling. The tool resamples with your browser’s high-quality smoothing, halving the image step by step first when it shrinks a lot, so neighbouring pixels are averaged rather than skipped. Making an image smaller is safe. Making it larger meets the pixel count but adds no detail, and Nature, Elsevier and others say so in their guidelines: re-export the figure from the original data or software at the size you need instead. For micrographs and blots, prefer the second option, keeping every pixel and only recording a new DPI, whenever the image already has enough pixels; resampling a data image interpolates new pixel values, and image-integrity guidance asks authors to keep and describe such processing (Cromey 2010).

Reading the DPI. PNG files record it in a pHYs chunk (pixels per metre), JPEG files in the JFIF header or EXIF data, and TIFF files in their XResolution and ResolutionUnit tags; a file with none recorded is usually shown at 72 or 96 dpi. The downloaded PNG gets a pHYs chunk with the chosen DPI, so Word, PowerPoint, InDesign and journal submission systems read the intended size.

Sources. Widths, resolutions and lettering are taken from each publisher’s current author guidelines, linked above, checked on 8 November 2026. Guidelines change and individual journals within a publisher can differ, so the target journal’s own page is the final word. Wiley titles set figure sizes journal by journal, so they are not listed. Cromey DW. Avoiding twisted pixels: ethical guidelines for the appropriate use and manipulation of scientific digital images. Sci Eng Ethics 2010;16:639–667, doi:10.1007/s11948-010-9201-y.

1. A two-panel figure for a Nature single column. A figure exported as 1800 × 1350 pixels, saved at 72 dpi (as screenshots and many plotting programs do):

1. The width: $89\mathrm{mm}÷25.4=3.504$ inches. At 300 dpi it needs $3.504\times 300=$ 1051 pixels across.
2. The image has 1800 pixels, so at 89 mm it would be $1800÷3.504=513.7$ dpi: enough. At its recorded 72 dpi it would be placed $1800÷72\times 25.4=635\mathrm{mm}$ wide, which is why it looks huge when inserted.
3. Resampled to the width: 1051 × 788 pixels ($1350\times 1051÷1800=788.25$), 66.75 mm tall, within Nature’s 170 mm limit. Or keep all 1800 × 1350 pixels and record 513.7 dpi: it is then placed 89 mm wide too.

2. A Western blot with labels for Cell Press, full width. $174\mathrm{mm}=6.850$ inches. As a greyscale image, 300 dpi: $6.850\times 300=$ 2055 pixels. If the journal treats it as line art, 1000 dpi: 6850 pixels. A 2400-pixel scan meets the first (350 dpi at 174 mm) but not the second; export the labels as vector text over the image, or ask the journal which category applies.

3. Line art for an Elsevier 1.5-column figure. $140\mathrm{mm}=5.512$ inches; at 1000 dpi that is 5512 pixels across, at 500 dpi (combination) 2756 pixels and at 300 dpi (halftone) 1654 pixels, the same values as Elsevier’s own table.

Text size. A letter of 7 pt is $7/72$ inch; at 300 dpi that is $7÷72\times 300=29$ pixels of font size, at 1000 dpi 97 pixels. If text in the image is smaller than that after resizing, it is below 7 pt in print: set the size of text in the plotting program at the final figure width rather than scaling the figure afterwards.