Nonlinear Chrominance Upsampling for YUV 4:2:0 Video Detail Preservation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video compression techniques, such as YUV 4:2:0, result in subsampling that attenuates high-frequency spatial content, leading to loss of detail in fine lines and text during upsampling, particularly in rendered video applications like CAD, where sharp lines become fuzzy due to averaging chrominance over multiple pixel blocks.

Innovation Solution

A computer-implemented method for upsampling compressed video that involves receiving an encoded video stream, identifying pixels, determining chrominance probabilities, and computing second chrominance values based on these probabilities to achieve a statistically more accurate representation of full resolution source video.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of energy

If YUV 4:2:0 subsampling is used to compress video data, then bandwidth consumption is reduced, but high-frequency spatial content is attenuated causing loss of detail in fine lines and text

Engineering Contradiction:
Improvebandwidth consumptionVSAvoidhigh-frequency spatial content
Core Design Contradiction:
Loss of energyVSLoss of information

Solution Approach 1:

The patent changes the chrominance calculation parameters by using probability-based nonlinear computation instead of traditional linear averaging. It calculates the probability that a second chrominance value equals the first chrominance value, then uses this probability to compute the second chrominance, thereby preserving high-frequency spatial information while maintaining compression benefits

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent replaces the traditional mechanical averaging process (low pass box filters) with a probabilistic computational approach. Instead of simply averaging chrominance values over pixel blocks, the system uses probability calculations to determine chrominance values, which better preserves edge information and fine details

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If traditional upsampling with averaging is used, then bandwidth is conserved, but sharp lines become fuzzy due to averaging chrominance over multiple pixel blocks

Engineering Contradiction:
Improvebandwidth efficiencyVSAvoidline definition accuracy
Core Design Contradiction:
ProductivityVSManufacturing precision

Solution Approach 1:

The patent fundamentally changes the upsampling parameter calculation by introducing probability-based nonlinear chrominance computation. Instead of uniform averaging, it calculates the probability that adjacent pixels share the same chrominance value and uses this probability to weight the chrominance reconstruction, thereby preserving sharp edges and fine line definitions

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent introduces dynamic probability-based chrominance calculation that adapts to local image characteristics. The system dynamically computes probabilities based on luminance differences and chrominance values, allowing the upsampling process to adaptively preserve edges in some regions while smoothing in others, thereby maintaining line definition accuracy

Inventive Principle:
Principle #15Dynamics

Data Source

PatentUS10250892B2Techniques for nonlinear chrominance upsampling
Publication Date: 2019.04.02 NVIDIA CORP
  • US10250892B2 patent drawing
  • US10250892B2 patent drawing
  • US10250892B2 patent drawing

AI summary

A subsystem configured to upsample a video data stream encoded in YCrCb format 4:2:0 (also termed YUV 4:2:0) performs an algorithm upon a two-by-two group of subsampled pixels. The subsystem computes an inside probability that the chrominance of a target pixel is a close match to the chrominance inside the group of four pixels. The subsystem further computes three weighting factors relating the chrominance of the target pixel to each of three adjacent pixels in an upsampled four-by-four pixel group. The subsystem then computes an outside estimate of the chrominance based on the weighting factors. Finally, the subsystem computes the chrominance of the target pixel based on the inside probability, the outside estimate, and the subsampled chrominance. The subsystem performs the algorithm upon all two-by-two groups of four pixels within a subsampled YUV 4:2:0 video data stream and generates an upsampled YUV 4:4:4 video data stream.