Lusamine.ImageIdentifier 1.3.0

Lusamine.ImageIdentifier

Identify an image's format and pixel dimensions by reading only its header. The full image is never loaded into memory. Works over any readable stream, including non-seekable ones (network, compression).

Supported formats: PNG, JPEG, GIF, BMP, WebP (VP8/VP8L/VP8X), TIFF, ICO.

Usage

using Lusamine.ImageIdentifier;

var identifier = new ImageIdentifier();

using var stream = File.OpenRead("photo.jpg");
ImageInfo? info = identifier.Identify(stream);

if (info is not null)
    Console.WriteLine($"{info.Format} {info.Width}x{info.Height}");

Identify returns null when the data is unrecognized or the header is truncated.

Why

Only the leading header bytes are read (typically well under 1 KB), so identifying a multi-megabyte image is allocation-light and near-instant. For non-seekable streams the reader skips forward by reading and discarding, so it still never buffers the whole file.

A few formats keep their dimensions past the first header bytes; these are reached by skipping forward, never by loading the file:

• JPEG — the SOFn marker is located by walking the marker chain, scanning up to 2 MB to clear large APPn metadata (a max-size EXIF block plus stacked ICC-profile segments).
• TIFF / BigTIFF — both the classic (magic 42, 32-bit offsets) and BigTIFF (magic 43, 64-bit offsets) variants are supported. The first IFD is read wherever it sits, up to a 128 MB offset cap. Backward IFD offsets are rejected since the reader is forward-only.
• SVG — width/height are read first (bare or px values), falling back to viewBox. Relative units (%, em, …) can't be resolved to pixels, so a file with only relative sizes and no viewBox returns null.

These caps bound work on crafted input; ordinary files resolve within the first read.

Performance

Benchmarks run against ImageSharp's Image.Identify on .NET 10, Windows 11, Intel Core i7-11800H:

And on .NET 10, macOS Tahoe 26.4, Apple M3:

Lusamine reads only the image header (typically under 1 KB), while ImageSharp decodes significantly more of the file. The result is consistently one to several orders of magnitude faster across formats, from ~11–13x for BMP up to hundreds of thousands of times faster for GIF, with a fraction of the allocations.

The advantage is just as large on the failure paths — unrecognized, empty, or corrupt input. Lusamine returns null cheaply and never allocates an exception, whereas ImageSharp throws and pays for it (.NET 10, macOS Tahoe 26.4, Apple M3):

Extensibility

The decoder set is injectable for custom formats or testing:

var identifier = new ImageIdentifier(new IImageFormatDecoder[] { new MyFormatDecoder() });