Implementation:Kornia Kornia Yuv Conversion

Knowledge Sources	Kornia
Domains	Vision, Color_Processing
Last Updated	2026-02-09 15:00 GMT

Overview

This module provides differentiable conversions between RGB and YUV color spaces, including full YUV 4:4:4, chroma-subsampled YUV 4:2:0, and YUV 4:2:2 formats.

Description

yuv.py is a module in the Kornia library's color subpackage implementing bidirectional RGB-to-YUV color space conversions. The YUV model follows M/PAL values as defined in ITU-R BT.470-5. Three resolution variants are provided:

• YUV 4:4:4: Full-resolution conversion via rgb_to_yuv / yuv_to_rgb using a 3x3 linear transformation matrix.
• YUV 4:2:0: Chroma-subsampled conversion via rgb_to_yuv420 / yuv420_to_rgb where UV planes are half-resolution in both dimensions.
• YUV 4:2:2: Chroma-subsampled conversion via rgb_to_yuv422 / yuv422_to_rgb where UV planes are half-resolution horizontally.

The output ranges are Y in (0, 1), U in (-0.436, 0.436), and V in (-0.615, 0.615). Each function has a corresponding nn.Module wrapper class.

Usage

Import this module when you need to convert between RGB and YUV color spaces, for example when working with video compression, broadcast standards, or luma-chroma separation in image processing pipelines.

Code Reference

Source Location

• Repository: Kornia
• File: kornia/color/yuv.py
• Lines: 1-470

Signature

def rgb_to_yuv(image: torch.Tensor) -> torch.Tensor: ...
def rgb_to_yuv420(image: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]: ...
def rgb_to_yuv422(image: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]: ...
def yuv_to_rgb(image: torch.Tensor) -> torch.Tensor: ...
def yuv420_to_rgb(imagey: torch.Tensor, imageuv: torch.Tensor) -> torch.Tensor: ...
def yuv422_to_rgb(imagey: torch.Tensor, imageuv: torch.Tensor) -> torch.Tensor: ...

class RgbToYuv(nn.Module): ...
class RgbToYuv420(nn.Module): ...
class RgbToYuv422(nn.Module): ...
class YuvToRgb(nn.Module): ...
class Yuv420ToRgb(nn.Module): ...
class Yuv422ToRgb(nn.Module): ...

Import

from kornia.color import rgb_to_yuv, yuv_to_rgb
from kornia.color import rgb_to_yuv420, yuv420_to_rgb
from kornia.color import rgb_to_yuv422, yuv422_to_rgb

I/O Contract

Inputs

Name	Type	Required	Description
image (rgb_to_yuv)	torch.Tensor	Yes	RGB image with shape (*, 3, H, W). Values in range (0, 1).
image (rgb_to_yuv420)	torch.Tensor	Yes	RGB image with shape (*, 3, H, W). H and W must be evenly divisible by 2.
image (rgb_to_yuv422)	torch.Tensor	Yes	RGB image with shape (*, 3, H, W). W must be evenly divisible by 2.
image (yuv_to_rgb)	torch.Tensor	Yes	YUV image with shape (*, 3, H, W). Y in (0,1), U in (-0.436, 0.436), V in (-0.615, 0.615).
imagey (yuv420_to_rgb)	torch.Tensor	Yes	Y plane with shape (*, 1, H, W). H and W must be evenly divisible by 2.
imageuv (yuv420_to_rgb)	torch.Tensor	Yes	UV planes with shape (*, 2, H/2, W/2).
imagey (yuv422_to_rgb)	torch.Tensor	Yes	Y plane with shape (*, 1, H, W).
imageuv (yuv422_to_rgb)	torch.Tensor	Yes	UV planes with shape (*, 2, H, W/2).

Outputs

Name	Type	Description
rgb_to_yuv return	torch.Tensor	YUV image with shape (*, 3, H, W).
rgb_to_yuv420 return	tuple[torch.Tensor, torch.Tensor]	Y plane (*, 1, H, W) and UV planes (*, 2, H/2, W/2).
rgb_to_yuv422 return	tuple[torch.Tensor, torch.Tensor]	Y plane (*, 1, H, W) and UV planes (*, 2, H, W/2).
yuv_to_rgb return	torch.Tensor	RGB image with shape (*, 3, H, W).
yuv420_to_rgb return	torch.Tensor	RGB image with shape (*, 3, H, W).
yuv422_to_rgb return	torch.Tensor	RGB image with shape (*, 3, H, W).

Usage Examples

Basic Usage

import torch
from kornia.color import rgb_to_yuv, yuv_to_rgb, rgb_to_yuv420, yuv420_to_rgb

# Full YUV 4:4:4 conversion
rgb = torch.rand(2, 3, 64, 64)
yuv = rgb_to_yuv(rgb)
print(yuv.shape)  # torch.Size([2, 3, 64, 64])

# Round-trip conversion
rgb_back = yuv_to_rgb(yuv)
print(rgb_back.shape)  # torch.Size([2, 3, 64, 64])

# YUV 4:2:0 with chroma subsampling
y_plane, uv_planes = rgb_to_yuv420(rgb)
print(y_plane.shape)   # torch.Size([2, 1, 64, 64])
print(uv_planes.shape) # torch.Size([2, 2, 32, 32])

# Reconstruct RGB from YUV 4:2:0
rgb_from_420 = yuv420_to_rgb(y_plane, uv_planes)
print(rgb_from_420.shape)  # torch.Size([2, 3, 64, 64])

# Using nn.Module wrapper
from kornia.color import RgbToYuv
converter = RgbToYuv()
yuv_out = converter(rgb)

Related Pages

• Principle:Kornia_Kornia_YUV_Color_Space