Merge pull request #62 from awxkee/dev

YCgCo improvements

Cargo.toml

@@ -2,7 +2,7 @@ workspace = { members = ["app", "fuzz"] }
 
 [package]
 name = "yuvutils-rs"
-version = "0.7.0"
+version = "0.8.0"
 edition = "2021"
 description = "High performance utilities for YUV format handling and conversion."
 readme = "README.md"
@@ -19,7 +19,7 @@ rust-version = "1.82.0"
 [dependencies]
 num-traits = "0.2.19"
 rayon = { version = "1.10.0", optional = true }
-fast_transpose = "0.2.3"
+fast_transpose = {  version = "0.2.3", optional = true }
 
 [dev-dependencies]
 rand = "0.9.0"
@@ -48,10 +48,11 @@ nightly_i8mm = []
 fast_mode = []
 # Enables `professional_mode` support on available paths
 professional_mode = []
-# Enables `rayon` support, use with care, in common, YUV encoding/decoding is more usually used in single thread mode
+# Enables `rayon` support, use with care, in common, YUV encoding/decoding is more usually expected to be used in single thread mode
 rayon = ["dep:rayon"]
 # Support for Big-Endian YUV
 big_endian = []
+geometry = ["dep:fast_transpose"]
 rdp = []
 
 [profile.dev.package]

app/benches/yuv8/main.rs

@@ -749,6 +749,34 @@ pub fn criterion_benchmark(c: &mut Criterion) {
         })
     });
 
+    c.bench_function("yuvutils YCgCo 4:2:2 Full -> RGBA", |b| {
+        let mut rgb_bytes = vec![0u8; dimensions.0 as usize * 4 * dimensions.1 as usize];
+        b.iter(|| {
+            yuv422_to_rgba(
+                &fixed_planar422,
+                &mut rgb_bytes,
+                dimensions.0 * 4u32,
+                YuvRange::Full,
+                YuvStandardMatrix::Bt601,
+            )
+            .unwrap();
+        })
+    });
+
+    c.bench_function("yuvutils YCgCo 4:2:2 Limited -> RGBA", |b| {
+        let mut rgb_bytes = vec![0u8; dimensions.0 as usize * 4 * dimensions.1 as usize];
+        b.iter(|| {
+            yuv422_to_rgba(
+                &fixed_planar422,
+                &mut rgb_bytes,
+                dimensions.0 * 4u32,
+                YuvRange::Limited,
+                YuvStandardMatrix::Bt601,
+            )
+            .unwrap();
+        })
+    });
+
     c.bench_function("libyuv YUV 4:2:2 -> RGBA", |b| {
         let mut rgb_bytes = vec![0u8; dimensions.0 as usize * 4 * dimensions.1 as usize];
         b.iter(|| unsafe {

app/src/main.rs

@@ -33,7 +33,17 @@ use image::{ColorType, DynamicImage, EncodableLayout, GenericImageView, ImageRea
 use std::fs::File;
 use std::io::Read;
 use std::time::Instant;
-use yuvutils_rs::{rdp_rgba_to_yuv444, rdp_yuv444_to_rgba, YuvBiPlanarImageMut, YuvChromaSubsampling, YuvPlanarImageMut};
+use yuvutils_rs::{
+    ar30_to_rgb8, i010_to_rgb10, i010_to_rgb_f16, i012_to_rgb12, i014_to_rgb14, i016_to_rgb16,
+    i210_to_rgb10, i210_to_rgb_f16, i210_to_rgba_f16, i214_to_rgb14, i214_to_rgb_f16,
+    i214_to_rgba14, i216_to_rgb16, i410_to_rgb10, i410_to_rgb_f16, i410_to_rgba10, i414_to_rgb14,
+    i414_to_rgb_f16, i416_to_rgb16, p210_to_ar30, p212_to_ar30, rgb10_to_i010, rgb10_to_i210,
+    rgb10_to_i410, rgb10_to_p210, rgb12_to_i012, rgb12_to_p212, rgb14_to_i014, rgb14_to_i214,
+    rgb14_to_i414, rgb16_to_i016, rgb16_to_i216, rgb16_to_i416, rgb_to_ycgco420, rgb_to_ycgco422,
+    rgb_to_ycgco444, rgba14_to_i214, rgba_to_ycgco420, ycgco420_to_rgb, ycgco420_to_rgba,
+    ycgco422_to_rgb, ycgco444_to_rgb, ycgco444_to_rgba, Rgb30ByteOrder, YuvBiPlanarImageMut,
+    YuvChromaSubsampling, YuvPlanarImageMut, YuvRange, YuvStandardMatrix,
+};
 
 fn read_file_bytes(file_path: &str) -> Result<Vec<u8>, String> {
     // Open the file
@@ -93,27 +103,39 @@ fn main() {
     );
 
     let mut planar_image =
-        YuvPlanarImageMut::<i16>::alloc(width as u32, height as u32, YuvChromaSubsampling::Yuv444);
+        YuvPlanarImageMut::<u8>::alloc(width as u32, height as u32, YuvChromaSubsampling::Yuv420);
     //
     let mut bytes_16: Vec<u16> = src_bytes
         .iter()
         .map(|&x| ((x as u16) << 4) | ((x as u16) >> 4))
         .collect();
 
     let start_time = Instant::now();
-    rdp_rgba_to_yuv444(&mut planar_image, src_bytes, rgba_stride as u32).unwrap();
+    rgba_to_ycgco420(
+        &mut planar_image,
+        &src_bytes,
+        rgba_stride as u32,
+        YuvRange::Full,
+    )
+    .unwrap();
 
     println!("Forward time: {:?}", start_time.elapsed());
     let fixed = planar_image.to_fixed();
     rgba.fill(0);
-    rdp_yuv444_to_rgba(&fixed, &mut rgba, rgba_stride as u32).unwrap();
 
     let fixed_biplanar = bi_planar_image.to_fixed();
     let fixed_planar = planar_image.to_fixed();
     // bytes_16.fill(0);
 
     let mut j_rgba = vec![0u8; dimensions.0 as usize * dimensions.1 as usize * 4];
 
+    ycgco420_to_rgba(
+        &fixed_planar,
+        &mut rgba,
+        dimensions.0 as u32 * 4,
+        YuvRange::Full,
+    )
+    .unwrap();
     // let mut rgba_f16: Vec<f16> = vec![0.; rgba.len()];
     // //
     // i210_to_rgb_f16(

fuzz/yuv_to_rgb/yuv_to_rgb.rs

@@ -31,9 +31,10 @@
 
 use libfuzzer_sys::fuzz_target;
 use yuvutils_rs::{
-    yuv420_alpha_to_rgba, yuv420_to_rgb, yuv420_to_rgba, yuv422_alpha_to_rgba, yuv422_to_rgb,
-    yuv422_to_rgba, yuv444_alpha_to_rgba, yuv444_to_rgb, yuv444_to_rgba, YuvPlanarImage,
-    YuvPlanarImageWithAlpha, YuvRange, YuvStandardMatrix,
+    ycgco420_to_rgb, ycgco420_to_rgba, ycgco422_to_rgb, ycgco422_to_rgba, yuv420_alpha_to_rgba,
+    yuv420_to_rgb, yuv420_to_rgba, yuv422_alpha_to_rgba, yuv422_to_rgb, yuv422_to_rgba,
+    yuv444_alpha_to_rgba, yuv444_to_rgb, yuv444_to_rgba, YuvPlanarImage, YuvPlanarImageWithAlpha,
+    YuvRange, YuvStandardMatrix,
 };
 
 fuzz_target!(|data: (u8, u8, u8, u8, u8, u8)| {
@@ -73,6 +74,14 @@ fn fuzz_yuv_420(i_width: u8, i_height: u8, y_value: u8, u_value: u8, v_value: u8
     )
     .unwrap();
 
+    ycgco420_to_rgb(
+        &planar_image,
+        &mut target_rgb,
+        i_width as u32 * 3,
+        YuvRange::Full,
+    )
+    .unwrap();
+
     let mut target_rgba = vec![0u8; i_width as usize * i_height as usize * 4];
 
     yuv420_to_rgba(
@@ -84,6 +93,14 @@ fn fuzz_yuv_420(i_width: u8, i_height: u8, y_value: u8, u_value: u8, v_value: u8
     )
     .unwrap();
 
+    ycgco420_to_rgba(
+        &planar_image,
+        &mut target_rgba,
+        i_width as u32 * 4,
+        YuvRange::Full,
+    )
+    .unwrap();
+
     let planar_image_with_alpha = YuvPlanarImageWithAlpha {
         y_plane: &y_plane,
         y_stride: i_width as u32,
@@ -139,6 +156,14 @@ fn fuzz_yuv_422(i_width: u8, i_height: u8, y_value: u8, u_value: u8, v_value: u8
     )
     .unwrap();
 
+    ycgco422_to_rgb(
+        &planar_image,
+        &mut target_rgb,
+        i_width as u32 * 3,
+        YuvRange::Full,
+    )
+    .unwrap();
+
     let mut target_rgba = vec![0u8; i_width as usize * i_height as usize * 4];
 
     yuv422_to_rgba(
@@ -150,6 +175,14 @@ fn fuzz_yuv_422(i_width: u8, i_height: u8, y_value: u8, u_value: u8, v_value: u8
     )
     .unwrap();
 
+    ycgco422_to_rgba(
+        &planar_image,
+        &mut target_rgba,
+        i_width as u32 * 4,
+        YuvRange::Full,
+    )
+    .unwrap();
+
     let planar_image_with_alpha = YuvPlanarImageWithAlpha {
         y_plane: &y_plane,
         y_stride: i_width as u32,

src/lib.rs

@@ -71,9 +71,11 @@ mod built_coefficients;
 mod f16_converter;
 mod from_identity;
 mod from_identity_alpha;
+#[cfg(feature = "geometry")]
 mod geometry;
 mod images;
 mod internals;
+#[cfg(feature = "geometry")]
 mod mirroring;
 #[cfg(all(target_arch = "aarch64", target_feature = "neon"))]
 mod neon;
@@ -272,9 +274,9 @@ pub use ycgco_to_rgb::{
 };
 
 pub use ycgco_to_rgb::{
-    icgc010_to_rgb, icgc010_to_rgba, icgc012_to_rgb, icgc012_to_rgba, icgc210_to_rgb,
-    icgc210_to_rgba, icgc212_to_rgb, icgc212_to_rgba, icgc410_to_rgb, icgc410_to_rgba,
-    icgc412_to_rgb, icgc412_to_rgba,
+    icgc010_to_rgb10, icgc010_to_rgba10, icgc012_to_rgb12, icgc012_to_rgba12, icgc210_to_rgb10,
+    icgc210_to_rgba10, icgc212_to_rgb12, icgc212_to_rgba12, icgc410_to_rgb10, icgc410_to_rgba10,
+    icgc412_to_rgb12, icgc412_to_rgba12,
 };
 
 pub use yuv_nv_to_rgba::yuv_nv16_to_bgr;
@@ -292,8 +294,8 @@ pub use ycgco_to_rgb_alpha::{
 };
 
 pub use ycgco_to_rgb_alpha::{
-    icgc010_alpha_to_rgba, icgc012_alpha_to_rgba, icgc210_alpha_to_rgba, icgc212_alpha_to_rgba,
-    icgc410_alpha_to_rgba, icgc412_alpha_to_rgba,
+    icgc010_alpha_to_rgba10, icgc012_alpha_to_rgba12, icgc210_alpha_to_rgba10,
+    icgc212_alpha_to_rgba12, icgc410_alpha_to_rgba10, icgc412_alpha_to_rgba12,
 };
 
 pub use yuv_to_yuy2::yuv420_to_uyvy422;
@@ -475,11 +477,12 @@ pub use f16_converter::{
     convert_rgba_f16_to_rgba16, convert_rgba_to_f16,
 };
 
+#[cfg(feature = "geometry")]
 pub use geometry::{
     rotate_cbcr, rotate_cbcr16, rotate_plane, rotate_plane16, rotate_rgb, rotate_rgb16,
     rotate_rgba, rotate_rgba16, RotationMode,
 };
-
+#[cfg(feature = "geometry")]
 pub use mirroring::{
     mirror_cbcr, mirror_cbcr16, mirror_plane, mirror_plane16, mirror_rgb, mirror_rgb16,
     mirror_rgba, mirror_rgba16, MirrorMode,

src/neon/mod.rs

@@ -69,6 +69,7 @@ mod utils;
 mod y_p16_to_rgba16;
 mod y_to_rgb;
 mod y_to_rgb_alpha;
+mod ycgco_to_rgb;
 mod yuv_nv_p10_to_ar30;
 mod yuv_nv_p10_to_rgba;
 #[cfg(feature = "professional_mode")]
@@ -161,6 +162,7 @@ pub(crate) use y_to_rgb::neon_y_to_rgb_row_rdm;
 pub(crate) use y_to_rgb_alpha::neon_y_to_rgb_alpha_row;
 #[cfg(feature = "rdm")]
 pub(crate) use y_to_rgb_alpha::neon_y_to_rgb_row_alpha_rdm;
+pub(crate) use ycgco_to_rgb::{neon_ycgco420_to_rgba_row, neon_ycgco_full_range_to_rgb};
 pub(crate) use yuv_nv_p10_to_ar30::neon_yuv_nv12_p10_to_ar30_row;
 pub(crate) use yuv_nv_p10_to_rgba::neon_yuv_nv12_p10_to_rgba_row;
 #[cfg(feature = "professional_mode")]