Implicit Masked Blending for Boundary-Aware Video Prediction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing video coding methods using compound prediction modes suffer from inaccuracies due to reference blocks extending beyond picture boundaries, leading to suboptimal padding and reduced prediction accuracy.

Innovation Solution

Implement an implicit masked blending mode that uses weighted averages of reference blocks within and outside boundary areas for more accurate video reconstruction, incorporating optical flow-based prediction and temporal interpolated prediction to enhance compound mode performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If reference blocks are used for compound prediction, then prediction coverage is improved, but prediction accuracy deteriorates due to out-of-boundary padding

Engineering Contradiction:

Improveprediction coverage areaVSAvoidprediction accuracy

Core Design Contradiction:

Area of stationary objectVSMeasurement precision

Solution Approach 1:

The patent applies different blending modes to different spatial regions: implicit masked blending for in-boundary areas and weighted average blending for out-of-boundary areas. This local differentiation ensures that each region uses the most appropriate blending strategy, maintaining high prediction accuracy while preserving adequate coverage area.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent introduces an implicit mask as an intermediary element that selectively controls the contribution of reference blocks based on their boundary status. This mask acts as a mediator between the conflicting requirements of coverage and accuracy, allowing the system to achieve both goals simultaneously through region-dependent blending.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Device complexity

If simple average blending is used for compound prediction, then computational complexity is reduced, but coding accuracy deteriorates due to artifacts

Engineering Contradiction:

Improveblending computation complexityVSAvoidcoding accuracy

Core Design Contradiction:

Device complexityVSManufacturing precision

Solution Approach 1:

The patent applies the more computationally intensive weighted average blending only partially - specifically for out-of-boundary regions - while using simpler implicit masked blending for in-boundary regions. This partial application of the heavier operation reduces overall computational complexity while maintaining coding accuracy where it matters most.

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The patent changes the blending parameter (weighting strategy) based on the spatial location and boundary status of reference blocks. By dynamically adjusting the blending parameters rather than using a fixed simple average, the system achieves higher coding accuracy without proportionally increasing computational complexity.

Inventive Principle:

Principle #35Parameter changes

3Area of stationary object

If out-of-boundary reference block portions are padded, then prediction coverage is maintained, but reconstruction accuracy deteriorates due to non-real pixel values

Engineering Contradiction:

Improvereference block coverageVSAvoidreconstruction accuracy

Core Design Contradiction:

Area of stationary objectVSManufacturing precision

Solution Approach 1:

The patent converts the harmful effect of out-of-boundary padding into a beneficial situation by detecting padded regions and applying a different blending mode (weighted average) specifically to those regions. This transforms the previously harmful padded areas into opportunities for improved accuracy through alternative processing.

Inventive Principle:

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

Solution Approach 2:

The patent performs preliminary identification of out-of-boundary regions before executing the blending operation. By pre-classifying which areas contain real pixel data and which contain padding, the system can apply appropriate blending strategies in advance, ensuring reconstruction accuracy is maintained throughout the process.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12477122B2Implicit masked blending mode improvement

Publication Date: 2025.11.18 TENCENT AMERICA LLC

AI summary

A computing system receives a video bitstream comprising a current picture and a current block within the current picture. The current block is encoded using information from a first reference block and a second reference block. When a portion of the current block corresponds to a first area that is within reference boundaries in both of the first and second reference blocks, the system reconstructs the portion using an average of reference values from the first and second reference blocks. When the portion of the current block corresponds to (i) a second area that is not within the reference boundaries in both of the first and second reference blocks or (ii) a third area that is within the reference boundaries in only one of the first and second reference blocks, the system reconstructs the portion using a weighted average of respective values of the first and second reference blocks.