Pixel Encoding Selective Denoising Saturated Color Artifacts

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing pixel encoding technologies face challenges in efficiently encoding pixels without significant deterioration of perceived quality, particularly due to severe artifacts caused by highly non-linear transfer functions and subsampling in saturated colors, which limit the color gamut and require additional bits for representation.

Innovation Solution

A method that selectively denoises color components with small variations that are not visually noticeable, converting them to a cheaper encoding format without affecting visual quality, using techniques like filtering and threshold-based modifications to reduce variability in Y'Cb'Cr' representation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If Y'Cb'Cr' color representation is used for encoding, then compression efficiency is improved, but severe artifacts appear in saturated colors due to non-linear transfer function and subsampling

Engineering Contradiction:
Improvecompression efficiencyVSAvoidartifacts in saturated colors
Core Design Contradiction:
ProductivityVSObject-affected harmful factors

Solution Approach 1:

The patent applies different processing strategies to different regions of the color space. Specifically, it identifies saturated color regions (where chroma components exceed thresholds) and applies artifact correction processing only to these regions, while leaving other regions unchanged. This localized approach maintains compression efficiency in normal regions while eliminating artifacts in saturated regions.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent modifies the encoding parameters for chroma components in saturated regions by adjusting the quantization matrices or applying offset corrections to Cb' and Cr' values. This changes the parameter representation to reduce the severity of artifacts while maintaining overall compression efficiency.

Inventive Principle:
Principle #35Parameter changes

2Object-affected harmful factors

If color gamut edges are avoided to prevent artifacts, then artifact quality is improved, but the range of reproducible colors is severely limited

Engineering Contradiction:
Improveartifact qualityVSAvoidcolor gamut range
Core Design Contradiction:
Object-affected harmful factorsVSAdaptability or versatility

Solution Approach 1:

Instead of avoiding saturated colors globally, the patent applies local correction only where needed. It processes saturated regions with artifact correction while preserving full color gamut coverage in non-saturated regions, thus maintaining both high color versatility and acceptable artifact quality.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent converts the harmful artifact-prone saturated regions into beneficial areas by applying specific correction techniques that actually improve the appearance of saturated colors. Rather than avoiding these regions, it uses them as targets for enhancement through chroma adjustment and selective processing.

Inventive Principle:
Principle #22Blessing in disguise (Convert harm into benefit)

3Object-affected harmful factors

If a transfer function with lower steepness is used to reduce artifacts, then artifact quality is improved, but more bits are required for representing each color component

Engineering Contradiction:
Improvebanding artifactsVSAvoidbits per color component
Core Design Contradiction:
Object-affected harmful factorsVSQuantity of substance

Solution Approach 1:

The patent applies different bit allocation strategies to different regions. In saturated regions where banding is problematic, it uses higher precision representation or adjusted quantization. In non-saturated regions, standard quantization is sufficient, thus optimizing overall bit usage while preventing banding artifacts where they matter most.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3298780B1Pixel processing and encoding
Publication Date: 2020.01.01 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
  • EP3298780B1 patent drawingFigure 1~2
  • EP3298780B1 patent drawingFigure 3~4
  • EP3298780B1 patent drawingFigure 5~7

AI summary

A method of processing pixels in a picture of a video sequence comprising multiple pictures comprises identifying a pixel to be processed in the picture for which a variation in a linear representation of a color of the pixel that is smaller than a first threshold value results in a variation in a non-linear representation of the color that is larger than a second threshold value. The variation in the linear representation of the color also results in a variation in a luminance and chrominance based representation of the color that is smaller than a third threshold value. The method also comprises processing the identified pixel by modifying a value of at least one color component of the pixel. The method achieves a selective denoising of particular color components in certain situations to reduce the encoding cost yet do not significantly affect the visual quality.