Rust image scale in different color spaces using SIMD and multithreading.

Supported NEON, SSE, AVX-2, AVX-512, AVX-VNNI, WASM.

This library provides for you a convenience to scale in different color spaces.
Prebuilt options for CIE L*a*b, CIE L*u*v, CIE L*c*h, Linear, Sigmoidal, Oklab, Jzazbz available.
Those transformations also very efficients. Prefer downscale in linear colorspace.
Upscaling might be done in LAB/LUB and simoidized components and also efficient in sRGB.

Have good f16 (the “binary16” type defined in IEEE 754-2008) support.

let img = ImageReader::open("./assets/asset.png") .unwrap() .decode() .unwrap(); let dimensions = img.dimensions(); let mut bytes = Vec::from(img.as_bytes()); let mut scaler = LinearScaler::new(ResamplingFunction::Lanczos3); scaler.set_threading_policy(ThreadingPolicy::Single); // ImageStore::<u8, 4> - (u8, 4) represents RGBA, (u8, 3) - RGB etc let store = ImageStore::<u8, 4>::from_slice(&mut bytes, dimensions.0 as usize, dimensions.1 as usize).unwrap(); let mut dst_store = ImageStoreMut::<u8, 4>::alloc(dimensions.0 as usize / 2, dimensions.1 as usize / 2); let plan = scaler.plan_rgba_resampling( ImageSize::new(dimensions.0 as usize, dimensions.1 as usize), //source size ImageSize::new(dimensions.0 as usize / 2, dimensions.1 as usize / 2), // target size true, // premultiply alpha ).unwrap(); plan.resample(&store, &mut dst_store).unwrap();

Despite all implementations are fast, not all the paths are implemented using SIMD, therefore some paths are slower. Here is a table which shows what's implemented with SIMD.

~ - Partially implemented

Features:

• To enable support of f16 the feature nightly_f16 should be activated and nightly compiler are required.
• nightly_avx512 activates AVX-512 feature set and requires nightly compiler channel.

For x86 and aarch64 NEON runtime dispatch is used.

neon optional target features are available, enable it when compiling on supported platform to get all features.

avx2, fma, sse4.1, f16c will be detected automatically if enabled, it will automatically detect and use the best path if enabled.

avx512 requires feature avx512 compiler channel, runtime detection if it is available then will be used.

avxvnni requires feature avx512, will be detected automatically if available, no additional actions need, it will automatically detect and use the best path if enabled. AVX-VNNI is helpful extension on modern Intel and AMD CPUs, consider turn it on to get maximum performance.

fullfp16, fhm NEON target detection performed in runtime, will be detected automatically if enabled, it will automatically detect and use the best path if enabled.

WASM simd128 target feature activating is mandatory in build flags.

To get full support of f16 nightly_f16 feature should be used. For NEON f16 feature use runtime detection, if CPU supports this feature then the very fast path is available

See picscale/include/picscale.h for more info

cd picscale && RUSTFLAGS="-C strip=symbols" cargo +nightly build -Z build-std=std,panic_abort --release

Over 30 resampling filters is supported.

Bilinear Nearest Cubic MitchellNetravalli CatmullRom Hermite BSpline Hann Bicubic Hamming Hanning Blackman

And others

This project is licensed under either of

• BSD-3-Clause License (see LICENSE)
• Apache License, Version 2.0 (see LICENSE)

at your option.