Pixel Pre-Processing for Chroma Subsampling Artifacts

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The combination of highly non-linear transfer functions and chroma subsampling in video processing leads to severe artifacts, particularly for saturated colors, due to mismatched interpolation of chroma and luma components, resulting in incorrect pixel intensities and visible distortions.

Innovation Solution

A pre-processing method that adjusts the luminance component (Y') in the YCbCr color space to minimize differences with the original linear luminance, using techniques such as binary search or look-up tables, to ensure accurate representation and reduce artifacts by compensating for errors in chroma components.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If chroma subsampling (4:2:0 or 4:2:2) is used to reduce data amount, then compression efficiency is improved, but severe artifacts appear in saturated colors due to mismatched interpolation of chroma and luma components

Engineering Contradiction:
Improvecompression efficiencyVSAvoidcolor accuracy
Core Design Contradiction:
ProductivityVSManufacturing precision

Solution Approach 1:

The patent applies a preliminary luminance correction step before chroma subsampling. It calculates a corrected luminance value by comparing the original luminance with the luminance that would result after chroma subsampling and upsampling. This correction compensates for the luminance shifts that will occur during subsequent chroma subsampling, preventing artifacts in saturated colors while maintaining compression efficiency.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements a feedback mechanism where the expected luminance after chroma subsampling is calculated and compared with the original luminance. The difference (error) is used to adjust the luminance value before subsampling. This closed-loop approach ensures that the final reconstructed image has accurate luminance values even after chroma subsampling and upsampling operations.

Inventive Principle:
Principle #23Feedback

2Measurement precision

If a highly non-linear transfer function is used to represent very high peak brightness, then luminance representation is improved, but banding artifacts appear in dark regions

Engineering Contradiction:
Improveluminance representationVSAvoiddark region quality
Core Design Contradiction:
Measurement precisionVSManufacturing precision

Solution Approach 1:

The patent applies different processing strategies to different luminance ranges. For pixels with very high peak brightness, the non-linear transfer function is used to maintain accurate luminance representation. For pixels in dark regions, the luminance correction mechanism ensures sufficient precision is maintained. This local adaptation prevents banding in dark regions while preserving the ability to represent very high peak brightness values.

Inventive Principle:
Principle #3Local quality

3Manufacturing precision

If 4:4:4 chroma sampling is used instead of 4:2:0 or 4:2:2, then color accuracy is improved, but data amount increases significantly

Engineering Contradiction:
Improvecolor accuracyVSAvoiddata amount
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The patent applies preliminary luminance correction before chroma subsampling, which compensates for the quality loss that would otherwise require 4:4:4 sampling. By pre-correcting the luminance values to account for upcoming chroma subsampling and upsampling operations, the patent maintains color accuracy comparable to 4:4:4 while achieving the compression benefits of 4:2:0 or 4:2:2 subsampling.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3257042B1Pixel pre-processing and encoding
Publication Date: 2022.06.22 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
  • EP3257042B1 patent drawingFigure 1~3
  • EP3257042B1 patent drawingFigure 4~5
  • EP3257042B1 patent drawingFigure 6~10

AI summary

A pixel pre-processing comprises obtaining an original linear luminance component value of a pixel in a picture in a third color space determined based on a linear color of the pixel in a first color space. A non-linear luma component value in a second color space is derived for the pixel based on a first non-linear chroma component value in the second color space, a second non-linear chroma component value in the second color space and the original linear luminance component value in the third color space. The pre-processing reduces luminance artifacts that otherwise may occur when chroma subsampling is used in combination with a non-linear transfer function.