[3.1.0] - 2025-01-09

• Updated version of fast_image_resize to 5.1.0.
• Added support for multi-thread image processing.
  You can create an instance of ResizerThreadPool and use it for set
  ResizeOptions.thread_pool field or put it as argument of AlphaMulDiv methods.
• Added support for optimization with help of SSE4.1 and AVX2 for
  the F32 pixel type.
• Added support for new pixel types F32x2, F32x3 and F32x4 with
  optimizations for SSE4.1 and AVX2.

[3.0.0] - 2024-05-15

• Updated version of fast_image_resize to 4.0.0.
• Added support of Neon CPU-instructions to speedup image resizing
  on ARM architecture.
• Added optional argument options: ResizeOptions into Resizer.resize()
  and Resizer.resize_pil() methods.
• Added support for images with zero sizes (width or/and height).
• Added FilterType.gaussian filter for convolution resize algorythm.
• Updated version of pyo3 to 0.21.2.
• BREAKING CHANGES:
  • Argument resize_alg was removed from Resizer.__init__() method.
    You have to use options argument of Resizer.resize()
    and Resizer.resize_pil() methods to change resize algorythm.
  • Resizer, by default, multiplies and divides color channels of image by
    an alpha channel. You may change this behavior through options argument.
  • Deleted support of Python 3.7.

[2.2.1] - 2024-02-15

• Fixed error with incorrect cropping of source image.

[2.2.0] - 2024-02-08

• Added support of Pillow >= 10.1.0.
• Updated version of pyo3 to 0.20.2.
• Updated version of fast_image_resize to 3.0.2.
• Added building of wheel for Python 3.12.

v2.1.2

• Added building of wheel for Python 3.11.
• Updated version of pyo3 to 0.17.2.

v2.1

• Added support of new types of pixels: U8x2, U16, U16x2 and U16x4.

v2.0

• Drop support for Python 3.6.
• Deleted variant sse2 from enum CpuExtensions.
• Added support of new types of pixels: U8x3 and U16x3.
• Added optimisation for convolution grayscale images (U8) with helps of SSE4.1 and AVX2 instructions.

v1.1

• Added support of new type of pixels - U8.
• Added support of L-type of PIL images (grayscale with one byte per pixel).

v1.0

First stable release.