Implementation:Kornia Kornia Ycbcr Conversion

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

Overview

This module provides differentiable conversion between RGB and YCbCr color spaces, as well as extraction of the luma (Y) channel.

Description

ycbcr.py is a module in the Kornia library's color subpackage implementing bidirectional RGB-to-YCbCr color space conversion. The conversion uses the ITU-R BT.601 luma coefficients (Y = 0.299*R + 0.587*G + 0.114*B) and derives Cb and Cr with a 0.5 offset to keep values positive. The module provides three functional APIs:

• rgb_to_ycbcr: Converts RGB to YCbCr (3-channel output).
• rgb_to_y: Extracts only the luma (Y) channel (1-channel output).
• ycbcr_to_rgb: Converts YCbCr back to RGB with output clamped to [0, 1].

Each conversion function (except rgb_to_y) has a corresponding nn.Module wrapper (RgbToYcbcr and YcbcrToRgb).

Usage

Import this module when you need to separate luma from chroma for JPEG-style processing, super-resolution (operating on the Y channel), or when working with video/image compression pipelines that use YCbCr.

Code Reference

Source Location

• Repository: Kornia
• File: kornia/color/ycbcr.py
• Lines: 1-177

Signature

def rgb_to_ycbcr(image: torch.Tensor) -> torch.Tensor: ...
def rgb_to_y(image: torch.Tensor) -> torch.Tensor: ...
def ycbcr_to_rgb(image: torch.Tensor) -> torch.Tensor: ...

class RgbToYcbcr(nn.Module):
    def forward(self, image: torch.Tensor) -> torch.Tensor: ...

class YcbcrToRgb(nn.Module):
    def forward(self, image: torch.Tensor) -> torch.Tensor: ...

Import

from kornia.color import rgb_to_ycbcr, ycbcr_to_rgb, rgb_to_y
from kornia.color import RgbToYcbcr, YcbcrToRgb

I/O Contract

Inputs

Name	Type	Required	Description
image (rgb_to_ycbcr)	torch.Tensor	Yes	RGB image with shape (*, 3, H, W). Values in range (0, 1).
image (rgb_to_y)	torch.Tensor	Yes	RGB image with shape (*, 3, H, W). Values in range (0, 1).
image (ycbcr_to_rgb)	torch.Tensor	Yes	YCbCr image with shape (*, 3, H, W). Values in range (0, 1).

Outputs

Name	Type	Description
rgb_to_ycbcr return	torch.Tensor	YCbCr image with shape (*, 3, H, W). Y (luma), Cb (blue-difference chroma), Cr (red-difference chroma).
rgb_to_y return	torch.Tensor	Luma channel with shape (*, 1, H, W).
ycbcr_to_rgb return	torch.Tensor	RGB image with shape (*, 3, H, W), clamped to [0, 1].

Usage Examples

Basic Usage

import torch
from kornia.color import rgb_to_ycbcr, ycbcr_to_rgb, rgb_to_y

# Convert RGB to YCbCr
rgb = torch.rand(1, 3, 128, 128)
ycbcr = rgb_to_ycbcr(rgb)
print(ycbcr.shape)  # torch.Size([1, 3, 128, 128])

# Extract only the luma (Y) channel
y_channel = rgb_to_y(rgb)
print(y_channel.shape)  # torch.Size([1, 1, 128, 128])

# Convert YCbCr back to RGB
rgb_back = ycbcr_to_rgb(ycbcr)
print(rgb_back.shape)  # torch.Size([1, 3, 128, 128])

# Super-resolution: operate on Y channel only
from kornia.color import RgbToYcbcr, YcbcrToRgb
to_ycbcr = RgbToYcbcr()
to_rgb = YcbcrToRgb()
ycbcr_out = to_ycbcr(rgb)
# Apply super-resolution model to ycbcr_out[:, 0:1, :, :]
# Then reconstruct: rgb_sr = to_rgb(ycbcr_sr)

Related Pages

• Principle:Kornia_Kornia_YCbCr_Color_Space