imagesize

This module analyzes JPEG/JPEG 2000/PNG/GIF/TIFF/SVG/Netpbm/WebP/BMP/AVIF/HEIC/HEIF image headers and returns image size, DPI, and related metadata.

import imagesize

width, height = imagesize.get("test.png")
print(width, height)

xdpi, ydpi = imagesize.getDPI("test.png")
print(xdpi, ydpi)

info = imagesize.get_info("test.png")
print(info.width, info.height, info.rotation, info.xdpi, info.ydpi, info.colors, info.channels)

This module is a pure Python module. You can use file like object like file or something like io.BytesIO.

Supported Python versions: 3.10-3.15

pip install imagesize

For local development setup:

python -m venv .venv
source .venv/bin/activate
pip install -e .

Version 2.0 includes the following updates:

• Added/expanded support for BMP.
• Added support for AVIF.
• Added support for HEIC/HEIF.
• imagesize.get_info() now returns richer metadata including color depth, channel count, and rotation information.
• Improved EXIF orientation handling so JPEG, TIFF, AVIF, and HEIC/HEIF return sizes that correctly reflect EXIF rotation metadata.
• Added type hints for the public API and related input/output types.

Backward incompatible behavior in 2.0:

• imagesize.get() now returns (-1, -1) when parsing fails.
• imagesize.getDPI() now returns (-1, -1) when parsing fails.

If your existing code relied on exceptions during parse failures, update it to explicitly check return values.

• imagesize.get(filepath: FileInput, *, exif_rotation: bool = True) -> tuple[int, int]

  Returns image size as (width, height). By default, orientation metadata is applied for rotated JPEG/TIFF/AVIF/HEIF images; pass exif_rotation=False to get the stored size as-is. On parsing errors it returns (-1, -1).

• imagesize.getDPI(filepath: FileInput) -> tuple[int, int]

  Returns image DPI as (xdpi, ydpi). On parsing errors it returns (-1, -1).

• imagesize.get_info(filepath: FileInput, *, size: bool = True, dpi: bool = True, colors: bool = True, exif_rotation: bool = True, channels: bool = True) -> ImageInfo

  Returns an ImageInfo named tuple with width, height, rotation, xdpi, ydpi, colors and channels fields. rotation contains orientation metadata (e.g. EXIF Orientation tag, or -1 when unavailable).

HTTP and HTTPS URLs are accepted as input. The library uses HTTP byte-range requests automatically so that pixel data does not need to be downloaded. If a server does not support range requests, it transparently falls back to one full download for compatibility. The first request is limited to 8 KiB; later random access expands cached regions in 64 KiB blocks only when more metadata is needed.

Only headers and the metadata needed for the requested result are read. Pixel payloads are skipped with seek() locally and HTTP Range requests remotely. Run the included benchmark with:

python bench.py --number 10000

The report includes elapsed time and the number of bytes read for PNG, JPEG, TIFF, GIF, JPEG2000, AVIF, and HEIF. It also simulates an 8 MiB remote AVIF and reports the number of Range requests and transferred bytes.

Run test with the following command:

python -m unittest

MIT License

• test/images/test.heic: https://nokiatech.github.io/heif/examples.html
• test/images/test.avif: https://libre-software.net/image/avif-test/

I referred to the following code:

• http://markasread.net/post/17551554979/get-image-size-info-using-pure-python-code
• https://stackoverflow.com/questions/8032642/how-to-obtain-image-size-using-standard-python-class-without-using-external-lib

I use sample image from here:

• https://www.nightprogrammer.org/development/multipage-tiff-example-download-test-image-file/

Thank you for feedback:

• tk0miya (https://latimes.pages.dev/https-github.com/tk0miya)
• shimizukawa (https://latimes.pages.dev/https-github.com/shimizukawa)
• xantares (https://latimes.pages.dev/https-github.com/xantares)
• Ivan Zakharyaschev (https://latimes.pages.dev/https-github.com/imz)
• Jon Dufresne (https://latimes.pages.dev/https-github.com/jdufresne)
• Geoff Lankow (https://latimes.pages.dev/https-github.com/darktrojan)
• Hugo (https://latimes.pages.dev/https-github.com/hugovk)
• Jack Cherng (https://latimes.pages.dev/https-github.com/jfcherng)
• Tyler A. Young (https://latimes.pages.dev/https-github.com/s3cur3)
• Mark Browning (https://latimes.pages.dev/https-github.com/mabrowning)
• ossdev07 (https://latimes.pages.dev/https-github.com/ossdev07)
• Nicholas-Schaub (https://latimes.pages.dev/https-github.com/Nicholas-Schaub)
• Nuffknacker (https://latimes.pages.dev/https-github.com/Nuffknacker)
• Hannes Römer (https://latimes.pages.dev/https-github.com/hroemer)
• mikey (https://latimes.pages.dev/https-github.com/ffreemt)
• Marco (https://latimes.pages.dev/https-github.com/marcoffee)
• ExtReMLapin (https://latimes.pages.dev/https-github.com/ExtReMLapin)
• gremur (https://latimes.pages.dev/https-github.com/gremur)
• fuyb1992 (https://latimes.pages.dev/https-github.com/fuyb1992)
• flagman (https://latimes.pages.dev/https-github.com/flagman)
• icemac (https://latimes.pages.dev/https-github.com/icemac)
• ojuuji (https://latimes.pages.dev/https-github.com/ojuuji)
• stephenfin (https://latimes.pages.dev/https-github.com/stephenfin)
• gaoflow (https://latimes.pages.dev/https-github.com/gaoflow)
• jdillard (https://latimes.pages.dev/https-github.com/jdillard)