Non-subsampled Video Encoding for H.264 Color Space Conversion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional image compression techniques, such as the H.264 standard, incur significant losses due to down sampling from R′G′B′ to Y′CrCb color space and subsequent encoding, leading to compounded data loss during image rendering and transmission.

Innovation Solution

Implementing non-subsampled encoding and decoding techniques by converting R′G′B′ data into separate Y′, Cb, and Cr planes, where the Y′-plane is set to be three times taller and the Cb and Cr planes are set to 128, using the H.264 standard for encoding and decoding, thereby avoiding subsampling losses.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If down sampling from R'G'B' to Y'CrCb color space is applied, then data volume is reduced, but image quality deteriorates due to subsampling losses

Engineering Contradiction:
Improvedata volumeVSAvoidimage quality
Core Design Contradiction:
Quantity of substanceVSManufacturing precision

Solution Approach 1:

The patent extracts only the essential luminance information (Y' plane) for compression while discarding the chrominance planes (Cr and Cb) that cause quality loss. By setting chroma_sampled_width and chroma_sampled_height to 1, the invention separates the useful luminance data from the problematic chrominance data, allowing compression without subsampling artifacts.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the sampling parameters of the Y'CrCb color space by setting chroma_sampled_width and chroma_sampled_height to 1, effectively converting from 4:2:2 or 4:2:0 subsampling to a non-subsampled representation. This parameter change allows the Y' plane to be processed at full resolution while minimizing the impact of chrominance data.

Inventive Principle:
Principle #35Parameter changes

2Productivity

If H.264 encoding is applied to compressed data, then transmission efficiency is improved, but compression loss increases

Engineering Contradiction:
Improvetransmission efficiencyVSAvoidcompression loss
Core Design Contradiction:
ProductivityVSLoss of information

Solution Approach 1:

The patent segments the Y'CrCb data into separate planes and applies different processing to each. The Y' plane is encoded with full resolution and appropriate compression settings, while the chrominance planes are set to minimum sampling (1x1). This segmentation allows the critical luminance information to be preserved while reducing overall data volume for efficient transmission.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different quality levels to different color planes: the Y' plane maintains high quality with full resolution encoding, while the Cr and Cb planes use minimal sampling. This local quality approach ensures that the most visually important information (luminance) is preserved while allowing compression of less critical chrominance data.

Inventive Principle:
Principle #3Local quality

3Manufacturing precision

If multiple color planes (Y', Cr, Cb) are encoded separately, then color information is preserved, but data complexity increases

Engineering Contradiction:
Improvecolor information preservationVSAvoiddata structure complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent merges the processing of multiple color planes into a unified encoding framework. By using the same H.264 encoding parameters for all planes and setting chroma_sampled_width/height to 1, the invention simplifies the data structure while maintaining separate color plane information. This merging approach reduces the complexity of handling multiple planes with different processing requirements.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentUS10034007B2Non-subsampled encoding techniques
Publication Date: 2018.07.24 NVIDIA CORP
  • US10034007B2 patent drawing
  • US10034007B2 patent drawing
  • US10034007B2 patent drawing

AI summary

Techniques for non-subsampled video encoding of R′G′B′ data using Y′, Cb and Cr data to generate compressed data wherein the Y′-plane comprises three separate color frames that are not interleaved, and recovering the data therefrom.