Video Coding Temporal Prediction Area Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video coding methods face challenges in achieving high coding efficiency while adhering to temporal prediction restrictions, particularly in scenarios where random access into a video stream can result in mismatches due to incorrect decoding of blocks in non-updated areas.

Innovation Solution

The method divides digitized images into specific image areas with associated reference areas, allowing temporal predictions only within these areas, using motion vectors to perform motion compensation and ensuring that reference blocks for prediction are exclusively based on pixel information from the corresponding reference area, with optional use of border extension algorithms for pixels outside the reference area.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If temporal prediction is performed using reference blocks from any area in the reference image, then coding efficiency is improved, but mismatches occur during random access due to incorrect decoding of blocks in non-updated areas

Engineering Contradiction:
Improvecoding efficiencyVSAvoiddecoding accuracy during random access
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The reference image is divided into multiple reference areas, and the current image is divided into multiple image areas. Each image area is associated with a specific reference area, and temporal prediction is restricted to use only pixel information from the associated reference area. This segmentation prevents mismatches during random access while maintaining coding efficiency through localized prediction.

Inventive Principle:
Principle #1Segmentation

2Reliability

If reference blocks are restricted to specific reference areas, then decoding accuracy during random access is improved, but coding efficiency decreases due to limited prediction options

Engineering Contradiction:
Improvedecoding accuracy during random accessVSAvoidcoding efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

Different reference areas are associated with different image areas based on their spatial and temporal characteristics. This local quality approach allows each region to use the most appropriate reference data, maintaining high coding efficiency while ensuring that random access decoding accuracy is preserved through area-specific prediction constraints.

Inventive Principle:
Principle #3Local quality

3Adaptability or versatility

If border extension algorithms are used for pixels outside reference areas, then the flexibility of reference block selection is improved, but processing complexity increases

Engineering Contradiction:
Improvereference block selection flexibilityVSAvoidprocessing complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

Border extension algorithms are applied selectively only to border reference blocks that extend beyond their associated reference area, rather than to all reference blocks. This partial action approach maintains flexibility in reference block selection while minimizing the increase in processing complexity by limiting the application of extension algorithms to only where necessary.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP2630800B1A method for coding a sequence of digitized images
Publication Date: 2019.09.11 SIEMENS AG
  • EP2630800B1 patent drawingFigure 1
  • EP2630800B1 patent drawingFigure 2~3
  • EP2630800B1 patent drawingFigure 4~5

AI summary

The invention refers to a method for coding a sequence of digitized images (I), wherein a motion compensation process is performed, said motion compensation process using motion vectors (MV) between image blocks (B1, B2, B3) referring to a number of pixels in a current image (I) and reference blocks (R1, R2, R3) referring to a number of pixels in a reference image (RI), said reference image (R1) being based on one or more images (I) out of the sequence. For each image block (B1, B2, B3) of at least a part of the current image (I) a temporal prediction based on a corresponding reference block (R1, R2, R3) indicated by a motion vector (MV) is performed, resulting in a prediction error between the image block (B1, B2, B3) and the corresponding reference block (R1, R2, R3), where said prediction error is coded. In the method of the invention, the current image (I) in the sequence is divided into several image areas (IA1, IA2, IA3, IA4 ), a reference area (RA1, RA2, RA3, RA4 ) in the reference image (RI) being associated with each image area (IA1, IA2, IA3, IA4 ), wherein the temporal prediction of an image block (Bl, B2, B3) in an image area (IA1, IA2, IA3, IA4 ) is based on a reference block (R1, R2, R3) at least partially located in the reference area (RA1, RA2, RA3, RA4 ) associated with the image area (IA1, IA2, IA3, IA4 ) and exclusively including pixel information from this reference area (RA1, RA2, RA3, RA4 ).