imagemeta

imagemeta is a high-performance Go library for extracting EXIF and XMP metadata from images and camera RAW formats.

It is built for:

• ExifTool-aligned decoding behavior
• low-allocation parsing on hot paths
• broad support for modern and legacy camera formats
• structured MakerNote decoding (Canon, Nikon, Sony, Panasonic, Apple)

Features

• Decode EXIF from JPEG, TIFF, CR2, CR3, DNG, NEF, ARW, RW2, HEIC/HEIF/AVIF, PNG, and CRW
• Parse XMP sidecars and embedded XMP blocks
• Read Canon/Nikon/Sony/Panasonic/Apple MakerNotes
• Extract embedded CR3 previews
• Fast image type detection
• Allocation-conscious perceptual image hashing (imagehash)

Install

go get github.com/evanoberholster/imagemeta

Supported Formats

• JPEG: .jpg, .jpeg
• TIFF + TIFF-based RAW: .tif, .tiff, .cr2, .dng, .nef, .arw, .rw2
• ISO-BMFF family: .cr3, .heic, .heif, .avif
• PNG (EXIF-in-TIFF payload)
• CIFF/CRW: .crw

Quick Start

package main

import (
    "fmt"
    "os"

    "github.com/evanoberholster/imagemeta"
)

func main() {
    f, err := os.Open("image.jpg")
    if err != nil {
        panic(err)
    }
    defer f.Close()

    ex, err := imagemeta.Decode(f)
    if err != nil {
        panic(err)
    }

    fmt.Println("type:", ex.ImageType)
    fmt.Println("make:", ex.Make)
    fmt.Println("model:", ex.Model)
    fmt.Println("datetime:", ex.DateTimeOriginal)
}

Decode API

Top-level decode entry points:

• Decode(io.ReadSeeker)
• DecodeJPEG(io.ReadSeeker)
• DecodeTiff(io.ReadSeeker)
• DecodeCR3(io.ReadSeeker)
• DecodeCRW(io.ReadSeeker)
• DecodeHeif(io.ReadSeeker)
• DecodePng(io.ReadSeeker)
• PreviewCR3(io.ReadSeeker)

XMP Parsing

Parse a sidecar .xmp file:

x, err := xmp.ParseXmp(file)

Parse embedded XMP from an image:

x, err := xmp.Parse(file)

Package: github.com/evanoberholster/imagemeta/meta/xmp

Image Type Detection

Use imagetype for quick format detection before decode:

t, err := imagetype.Scan(r)

Package: github.com/evanoberholster/imagemeta/imagetype

Perceptual Hashing

imagehash provides 64-bit and 256-bit perceptual hashing with low-allocation paths.

Package: github.com/evanoberholster/imagemeta/imagehash

Performance Notes

• Designed for low allocation and high throughput
• Uses pooled readers and fixed-size parsing structures on hot paths
• Suitable for large-batch metadata extraction pipelines

Benchmarks

• General benchmarks: bench_test.go

Note: benchmark results vary by CPU, Go version, and sample files.

TODO

• Expand ExifTool parity coverage across MakerNote fields (Nikon, Sony, Panasonic, Apple, DJI)
• Resolve known reverse-offset parsing edge case in MakerNote decoding
• Improve malformed metadata resilience with more targeted fuzz and corpus tests
• Add more specific benchmarks to compare edge cases
• Document compatibility matrix by camera model/format

Contributing

Pull requests and issue reports are welcome.

Before opening a PR, run:

go test ./...
golangci-lint run

Acknowledgements

imagemeta is heavily inspired and tested on:

• Phil Harvey's ExifTool: https://exiftool.org
• rwcarlsen/goexif

Additional references:

• go4 (BMFF/HEIF parser concepts)
• lclevy/canon_cr3
• Lasse Heikkilä's HEIF thesis work

LLMs were used to assist in writing code, the code has been reviewed and has passed stringent checks.

License

MIT