Video Coding Apparatus Optimizing Read Data Volume

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional frame coding techniques experience increased read data volumes and processing delays due to reading data specified by reduction motion vectors and plural motion vectors.

Innovation Solution

A coding apparatus that identifies the member count of a vector group specifying search positions on a reference frame and determines a maximum read data volume per unit time to optimize data reading for motion vector detection, reducing the search range and improving coding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If data of search range specified by reduction motion vector is read together with data of search range specified by plural motion vectors, then motion vector detection accuracy is improved, but read data volume increases and processing delays occur

Engineering Contradiction:
Improvemotion vector detection accuracyVSAvoidprocessing delay
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies partial action by selectively reading only the necessary portion of reference frame data. Instead of reading all search range data specified by reduction motion vectors together with plural motion vectors, the system determines a maximum read data volume and selectively reads data based on this constraint, performing partial reading action to avoid excessive data processing while maintaining adequate motion vector detection accuracy

Inventive Principle:
Principle #16Partial or excessive action

2Measurement precision

If data of search range specified by reduction motion vector is read together with data of search range specified by plural motion vectors, then motion vector detection accuracy is improved, but read data volume increases

Engineering Contradiction:
Improvemotion vector detection accuracyVSAvoidread data volume
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent changes the parameter of read data volume by introducing a maximum read data volume constraint. The system determines this maximum volume based on memory bandwidth capabilities and uses it to control the amount of reference frame data read, thereby adjusting the data volume parameter to prevent memory bandwidth exhaustion while maintaining adequate motion vector detection accuracy

Inventive Principle:
Principle #35Parameter changes

3Reliability

If maximum read data volume is determined based on memory bandwidth and vector group member count, then memory bandwidth exhaustion is prevented, but search range is reduced

Engineering Contradiction:
Improvememory bandwidth managementVSAvoidsearch range area
Core Design Contradiction:
ReliabilityVSArea of stationary object

Solution Approach 1:

The patent applies dynamics by making the search range adaptive rather than fixed. The search range is dynamically adjusted based on the determined maximum read data volume and vector group member count. This allows the search range to flexibly expand or contract according to memory bandwidth availability and coding requirements, preventing memory bandwidth exhaustion while maintaining adequate search coverage

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP2981089B1Encoding device, computer system, encoding method, encoding program, and recording medium
Publication Date: 2019.07.10 FUJITSU LTD
  • EP2981089B1 patent drawingFigure 1(a)~1(b)
  • EP2981089B1 patent drawingFigure 2
  • EP2981089B1 patent drawingFigure 3

AI summary

Delays of processing involved in frame coding are suppressed and coding efficiency is improved. A coding apparatus (100) identifies 3 as the member count of a vector group including a reduction motion vector (MVrt), a prediction vector (MVpt1), and a prediction vector (MVpt2). The coding apparatus (100) determines based on the identified count and a maximum read data volume (Ssum), a first maximum read data volume (SO) for a single given frame when image data of a search range that includes search positions specified on a reference frame (rF1) by the reduction motion vector is read in.