Fixed #

• Android 15 compatibility - Added 16KB page alignment for native library
  • Fixed libflutter_bicubic_resize.so alignment from 4KB to 16KB
  • Required for Android 15 (API 35+) which enforces 16KB page size
  • Added -Wl,-z,max-page-size=16384 linker flag in CMakeLists.txt

Added #

• ML preprocessing with normalization - new resizeForModel() method returning Float32List
  • NormalizationType enum: none (default), simple [0,1], centered [-1,1], imageNet, custom
  • ChannelOrder enum: rgb (default), bgr
  • TensorLayout enum: hwc (default, TensorFlow), chw (PyTorch)
  • Custom mean/std normalization parameters
  • Native C pipeline for raw RGB output (no re-encoding overhead)

Optimized #

• Pre-computed scale/offset factors for normalization (multiply+add instead of divide+subtract+divide)
• Branch conditions moved outside hot loops for better performance

Changed #

• Added machine-learning topic to pubspec.yaml
• Updated API documentation with ML preprocessing section
• Updated README with normalization examples

Added #

• Format detection and validation - library now handles unsupported formats
  • ImageFormat enum (jpeg, png) for supported formats
  • UnsupportedImageFormatException - thrown for unsupported formats (HEIC, WebP, GIF, etc.)
  • BicubicResizer.detectFormat(bytes) - detect image format from bytes
  • BicubicResizer.resize(bytes: ...) - generic resize with auto-detection, throws exception for unsupported formats

Added #

• Crop anchor positions (CropAnchor) - crop from any position, not just center
  • center (default), topLeft, topCenter, topRight, centerLeft, centerRight, bottomLeft, bottomCenter, bottomRight
• Crop aspect ratio modes (CropAspectRatio) - control crop shape
  • square (default) - 1:1 aspect ratio
  • original - keep original image proportions
  • custom - use custom aspect ratio with aspectRatioWidth/aspectRatioHeight
• Edge handling modes (EdgeMode) - control how pixels outside bounds are handled
  • clamp (default) - repeat edge pixels
  • wrap - tile/repeat image
  • reflect - mirror reflection at edges
  • zero - black/transparent pixels outside
• JPEG EXIF control (applyExifOrientation) - option to disable EXIF orientation correction
• PNG compression control (compressionLevel) - adjust compression 0-9 (default: 6)

Changed #

• All new parameters are optional with backward-compatible defaults
• Updated API documentation with comprehensive examples
• Expanded README with new features and usage examples

Fixed #

• iOS release build crash - Fixed FFI symbol stripping issue that caused "symbol not found" errors in release/archive builds
• Added proper Xcode build settings to prevent symbol stripping (STRIP_STYLE=non-global)
• Added symbol retention mechanism in Swift plugin to ensure C functions are linked

Added #

• EXIF orientation support for JPEG images - photos from mobile cameras now display correctly
• Automatic rotation/flip based on EXIF metadata (supports all 8 orientation values)

Changed #

• Breaking: Crop now produces 1:1 aspect ratio (square) instead of proportional crop
  • crop: 1.0 on 1920x1080 image now crops to 1080x1080 (not 1920x1080)
  • This prevents stretching when resizing to square output (e.g., 224x224)
• Updated README with new features documentation

Added #

• Optional center crop parameter (crop) for all resize methods
• Crop value range: 0.0-1.0 (1.0 = no crop, 0.5 = center 50%)
• Crop is applied before resize for efficient single-pass processing
• Example app now includes crop slider for testing

Changed #

• Updated API documentation with crop parameter
• Updated README with crop usage examples

Fixed #

• Fixed GitHub repository links in pubspec.yaml

Added #

• Added API documentation (doc/api.md)
• Added topics for better discoverability on pub.dev
• Added funding link
• Added library documentation comments

Added #

• Initial release
• Bicubic resize using native C code (stb_image_resize2)
• RGB and RGBA support
• JPEG and PNG encoding/decoding
• Isolate support for non-blocking operations
• iOS and Android support