Rust utilities for YUV format handling and conversion.

Fast and simple YUV approximation conversion in pure Rust. At most the same as libyuv does. Performance will be equal to libyuv or slightly higher on platforms where SIMD is implemented. Otherwise, equal or slower. Used precision is slightly higher than used in libyuv, however for decoding 4:2:0 is almost always faster than libyuv, decoding 4:2:2, 4:4:4 almost the same, however encoding except 4:2:0 often is slower with better results. Modern NEON CPU is almost always faster than libyuv.

Mostly implemented AVX-512, AVX2, SSE, NEON, WASM. Waiting for SVE to be available in nightly.

x86 targets with SSE and AVX uses runtime dispatch to detect available cpu features.

Supports:

• YCbCr ( aka YUV )
• YCgCo
• YUY2
• Identity ( GBR )
• Sharp YUV

All the methods support RGB, BGR, BGRA and RGBA

SIMD

Runtime dispatch is used for use if available sse4.1, avx2, avx512bw, avx512vbmi, avx512vnni, avxvnni, rdm for NEON.

For the sse4.1, avx2, rdm there is no action needed, however for AVX-512 activated feature nightly_avx512 along nightly rust channel is required.

Always consider to compile with nightly_avx512 when compiling for x86, it adds noticeable gain on supported cpus.

Wasm simd128 as target feature should be enabled for implemented SIMD wasm paths support,

Rayon

Some paths have multi-threading support, consider this feature if you're working on platform where using of multi-threading is reasonable.

Adding to project

cargo add yuvutils-rs

RGB to YCbCr

let mut planar_image =
    YuvPlanarImageMut::<u8>::alloc(width as u32, height as u32, YuvChromaSubsampling::Yuv420);
rgb_to_yuv422(
    &mut planar_image,
    &src_bytes,
    rgba_stride as u32,
    YuvRange::Limited,
    YuvStandardMatrix::Bt601,
)
.unwrap();

YCbCr to RGB

yuv420_to_rgb(
    &yuv_planar_image,
    &mut rgba,
    rgba_stride as u32,
    YuvRange::Limited,
    YuvStandardMatrix::Bt601,
)
.unwrap();

Benchmarks

Tests performed on the image 1997x1331

YUV 8 bit-depth conversion

aarch64 tested on Mac Pro M3.

AVX2 tests performed on m4.large with Ubuntu 24.04 droplet. AVX2 Win test performed on Windows 11 Intel Core i9-14900HX.

cargo bench --bench yuv8 --manifest-path ./app/Cargo.toml

AVX-512 tests performed on AWS c5.large with Ubuntu 24.04 instance with command

cargo +nightly bench --bench yuv8 --manifest-path ./app/Cargo.toml --features nightly_avx512

Encoding 8-bit

Conversion	time(NEON)	Time(AVX2 Win)	Time(AVX2)	Time(AVX-512)
utils RGB->YUV 4:2:0	360.53µs	467.74µs	1.05ms	722.87µs
libyuv RGB->YUV 4:2:0	306.80µs	3.70ms	1.10ms	972.52µs
utils RGBA->YUV 4:2:0	422.36µs	447.70µs	1.09ms	785.93µs
libyuv RGBA->YUV 4:2:0	429.36µs	2.70ms	808.11µs	707.33µs
utils RGBA->YUV 4:2:2	590.36µs	567.99µs	1.27ms	964.87µs
libyuv RGBA->YUV 4:2:2	651.84µs	4.3003ms	1.09ms	947.97µs
utils RGBA->YUV 4:4:4	674.61µs	597.30µs	1.57ms	974.29µs

Decoding 8-bit

Conversion	time(NEON)	Time(AVX2 Win)	Time(AVX2)	Time(AVX-512)
utils YUV NV12->RGBA	399.37µs	377.72µs	954.72µs	736.21µs
utils YUV NV12->RGB	470.35µs	387.70µs	895.20µs	654.37µs
libyuv YUV NV12->RGB	621.11µs	5.5533ms	979.96µs	800.85µs
utils YUV 4:2:0->RGB	346.13µs	375.34µs	946.66µs	614.19µs
libyuv YUV 4:2:0->RGB	747.18µs	5.8231ms	1.42ms	1.16ms
utils YUV 4:2:0->RGBA	396.58µs	380.13µs	954.50µs	733.88µs
libyuv YUV 4:2:0->RGBA	710.88µs	557.44µs	1.03ms	870.64µs
utils YUV 4:2:2->RGBA	471.54µs	410.71µs	1.10ms	824.69µs
libyuv YUV 4:2:2->RGBA	734.25µs	579.24µs	1.04ms	851.24µs
utils YUV 4:4:4->RGBA	458.71µs	366.85µs	1.13ms	819.17µs
libyuv YUV 4:4:4->RGBA	596.55µs	525.90µs	963.66µs	803.62µs
utils YUV 4:0:0->RGBA	220.67µs	254.83µs	676.04µs	600.43µs
libyuv YUV 4:0:0->RGBA	522.71µs	1.8204ms	635.61µs	525.56µs

YUV 16 bit-depth conversion

cargo bench --bench yuv16 --manifest-path ./app/Cargo.toml

AVX-512 tests performed on AWS c5.large with Ubuntu 24.04 instance with command

cargo +nightly bench --bench yuv16 --manifest-path ./app/Cargo.toml --features nightly_avx512

Encoding 10-bit

10-bit encoding is not implemented in libyuv

	time(NEON)	Time(AVX2 Win)	Time(AVX2)	Time(AVX-512)
utils RGB10->YUV10 4:2:0	539.82µs	745.02µs	1.88ms	1.21ms
libyuv RGB10->YUV10 4:2:0	x	x	x	x
utils RGBA10->YUV10 4:2:0	674.12µs	807.89µs	2.03ms	1.57ms
libyuv RGBA10->YUV10 4:2:0	x	x	x	x
utils RGBA10->YUV10 4:2:2	838.82µs	1.04ms	2.65ms	1.83ms
libyuv RGBA10->YUV10 4:2:2	x	x	x	x
utils RGBA10->YUV10 4:4:4	659.93µs	1.12ms	3.07ms	2.03ms

Decoding 10-bit

	time(NEON)	Time(AVX2 Win)	Time(AVX2)	Time(AVX-512)
utils YUV10 4:2:0->RGB10	640.34µs	662.38µs	1.78ms	1.23ms
libyuv YUV10 4:2:0->RGB10	x	x	x	x
utils YUV10 4:2:0->RGBA10	814.86µs	670.15µs	1.97ms	1.59ms
utils YUV10 4:2:0->RGBA8	812.53µs	492.63µs	1.36ms	900.10µs
libyuv YUV10 4:2:0->RGBA8	1.7037ms	6.8641ms	1.14ms	966.94µs
utils YUV10 4:2:2->RGBA10	859.39µs	787.41µs	2.02ms	1.61ms
utils YUV10 4:2:2->RGBA8	878.54µs	438.14µs	1.23ms	972.19µs
libyuv YUV10 4:2:2->RGBA8	1.7056ms	6.8374ms	1.21ms	991.23µs
utils YUV10 4:4:4->RGBA10	931.28µs	848.02µs	2.34ms	1.8639ms

This project is licensed under either of

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

at your option.