flutter_opencv_ffi

Dart FFI bindings to OpenCV for image processing. Provides a high-level CvImage class for common operations: load/save, color conversion, blur, edge detection, and resize.

Uses Dart build hooks to compile a thin C++ wrapper at build time — no manual native setup required.

Platform Support

Platform	Status	OpenCV Source
macOS	Supported	Homebrew (brew install opencv)
iOS	Supported	Auto-downloaded pre-built framework
Android	Supported	Auto-downloaded pre-built SDK
Windows	Not yet	—
Linux	Not yet	—

Requirements

• Dart SDK >= 3.11.0
• macOS: OpenCV installed via Homebrew (brew install opencv)
• iOS / Android: No extra setup — the build hook downloads pre-built OpenCV SDKs automatically.

Installation

Add to your pubspec.yaml:

dependencies:
  flutter_opencv_ffi:
    git:
      url: https://github.com/enspyrco/flutter_opencv_ffi.git

Usage

import 'package:flutter_opencv_ffi/flutter_opencv_ffi.dart';

// Check OpenCV version.
print(opencvVersion()); // e.g. "4.13.0"

// Load an image.
final image = CvImage.imread('photo.jpg');

// Convert to grayscale.
final gray = image.cvtColor(ColorConversion.bgrToGray);

// Apply Gaussian blur.
final blurred = gray.gaussianBlur(ksize: 5, sigmaX: 1.5);

// Detect edges.
final edges = blurred.canny(threshold1: 50, threshold2: 150);

// Resize.
final thumbnail = image.resize(width: 100, height: 100);

// Encode to PNG bytes.
final pngBytes = edges.imencode('.png');

// Save to disk.
edges.imwrite('edges.png');

// Clean up native memory.
edges.dispose();
blurred.dispose();
gray.dispose();
thumbnail.dispose();
image.dispose();

In-memory round-trip

import 'dart:typed_data';

// Decode from bytes (e.g. from network or assets).
final Uint8List jpegData = fetchImageBytes();
final image = CvImage.imdecode(jpegData);

// Process and re-encode.
final processed = image.gaussianBlur(ksize: 3, sigmaX: 0);
final outputBytes = processed.imencode('.jpg');

processed.dispose();
image.dispose();

API Reference

Top-level

Function	Description
opencvVersion()	Returns the OpenCV version string

CvImage

Constructor / Factory	Description
CvImage.imread(path, {flags})	Load image from file
CvImage.imdecode(data, {flags})	Decode image from byte buffer

Property	Description
rows	Image height
cols	Image width
channels	Number of channels (e.g. 3 for BGR)
type	OpenCV Mat type code
isEmpty	Whether the Mat has no data

Method	Description
clone()	Deep copy
imwrite(path)	Save to file
imencode(ext)	Encode to byte buffer
cvtColor(conversion)	Color space conversion
gaussianBlur(ksize:, sigmaX:, sigmaY:)	Gaussian blur
canny(threshold1:, threshold2:)	Canny edge detection
resize(width:, height:, interpolation:)	Resize image
dispose()	Release native memory

Enums

• ColorConversion — bgrToGray, grayToBgr, bgrToRgb, rgbToBgr, bgrToRgba, rgbaToBgr
• Interpolation — nearest, linear, cubic, area, lanczos4

Architecture

Dart (CvImage)  -->  @Native bindings (ffigen)  -->  C wrapper (extern "C")  -->  OpenCV C++

The C wrapper (src/opencv_wrapper.h / .cpp) provides a flat C API using opaque CvMat* handles. ffigen generates @Native annotated Dart bindings. The build hook (hook/build.dart) compiles the C++ wrapper and links it against OpenCV at build time.

Development

# Install OpenCV (macOS)
brew install opencv

# Get dependencies
dart pub get

# Regenerate FFI bindings after modifying src/opencv_wrapper.h
dart run ffigen --config ffigen.yaml

# Run tests
dart test

# Run the example app
cd example && flutter run

License

See LICENSE for details.