Video Coding Pre-Processing for CTU Boundary Prediction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Massively parallel video coding (MPC) incurs coding efficiency losses due to prohibited coding dependencies among coding tree units (CTUs) since reference samples from adjacent reconstructed blocks are not available for intra prediction, especially in high-resolution content and large-scale video streaming.

Innovation Solution

Implement pre-processing operations based on directional correlation to reduce the number of samples along strong correlation directions, including downsampling, packing, and applying one-dimensional spatial-to-frequency domain transforms on image samples, while signaling reference line directions and limiting intra sub-partitioning to enhance coding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If CTUs are coded independently in massively parallel coding, then parallel processing capability is improved, but coding efficiency deteriorates due to loss of reference samples

Engineering Contradiction:

Improveparallel processing capabilityVSAvoidcoding efficiency

Core Design Contradiction:

ProductivityVSLoss of information

Solution Approach 1:

The picture is divided into multiple independent CTUs that can be processed in parallel. Each CTU is a self-contained unit with its own processing pipeline, allowing simultaneous encoding/decoding of multiple CTUs without inter-dependency, thus achieving massive parallel processing capability.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Reference line data structures are introduced as intermediaries to store and manage reference samples from adjacent CTUs. These data structures act as mediators that enable information exchange between independently processed CTUs, allowing reference samples to be available for intra prediction even when CTUs are coded in parallel.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Productivity

If reference samples are not shared among CTUs, then parallel coding is enabled, but intra prediction accuracy deteriorates

Engineering Contradiction:

Improveparallel coding capabilityVSAvoidintra prediction accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

Reference line data structures are pre-initialized and populated with reference samples from adjacent CTUs before the current CTU is processed. This preliminary action ensures that when intra prediction is performed on the current CTU, the necessary reference samples are already available, maintaining prediction accuracy without compromising parallel processing.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The reference line data structures serve as intermediary storage that bridges independently processed CTUs. They capture and preserve reference sample information from neighboring CTUs, making this information available for intra prediction operations while maintaining the independence of parallel CTU processing.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Reliability

If all image samples are processed with full detail, then image quality is maintained, but computational complexity increases

Engineering Contradiction:

Improveimage qualityVSAvoidcomputational complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

Different processing approaches are applied to different regions of the image based on local characteristics. Reference lines and boundary samples receive special treatment with dedicated data structures, while interior samples use standard processing. This local differentiation maintains image quality at boundaries while reducing overall computational complexity.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

Reference line data structures extract and separate the boundary and reference samples from the main image processing pipeline. By isolating these critical samples into dedicated structures, the patent reduces the computational burden on the main processing path while ensuring high-quality handling of boundary regions that are essential for parallel processing correctness.

Inventive Principle:

Principle #2Taking out (Extraction)

Data Source

PatentUS12621455B2Apparatus, a method and a computer program for video coding and decoding

Publication Date: 2026.05.05 NOKIA TECHNOLOGIES OY

AI summary

A method comprising: obtaining image blocks (702) comprising image samples (704, 706); performing one or more pre-processing operations for at least part of the image samples (704, 706) based on directional correlation in said samples (704, 706); and encoding the pre-processed image samples (704, 706) with video or image coding means.