Forward Motion Compensation Block Partitioning for Image Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image coding techniques using forward motion compensation struggle to accurately assign a single motion vector to blocks covered by multiple projected blocks, leading to suboptimal prediction and increased coding costs due to the use of approximate motion trajectories.

Innovation Solution

The method involves partitioning current blocks into sub-blocks and iteratively applying a predetermined allocation criterion to select the most suitable projected motion vector, ensuring that each sub-block is covered by a single motion vector, thereby optimizing motion compensation and reducing coding costs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If forward motion compensation is used to project blocks from reference image to current image, then motion prediction accuracy is improved, but overlap zones with multiple motion vectors and uncovering zones without motion vectors are created

Engineering Contradiction:
Improvemotion prediction accuracyVSAvoidmotion vector assignment complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The current image is divided into multiple current blocks, and each current block is independently processed to determine its motion vector. This segmentation allows the system to handle overlap and uncovering zones by treating each block separately, selecting or computing appropriate motion vectors for each region without being constrained by the global overlap/uncovering issues.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different motion vector assignment strategies are applied to different current blocks based on their local characteristics. Blocks in overlap zones use one approach (e.g., selecting from multiple candidate vectors), while blocks in uncovering zones use another approach (e.g., interpolation or default motion vectors), optimizing prediction accuracy locally for each region.

Inventive Principle:
Principle #3Local quality

2Ease of manufacture

If a single motion vector is assigned to blocks covered by multiple projected blocks, then coding simplicity is improved, but motion prediction accuracy deteriorates due to inability to represent complex motion trajectories

Engineering Contradiction:
Improvecoding simplicityVSAvoidmotion trajectory accuracy
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The system dynamically selects the appropriate motion vector for each current block based on its specific characteristics and the characteristics of covering projected blocks. This dynamic selection process allows the system to maintain coding simplicity (single vector per block) while adapting to local motion variations, thereby preserving prediction accuracy.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The motion vector parameters (horizontal and vertical components) are adjusted based on the specific conditions of each current block and its covering projected blocks. By changing these parameters adaptively, the system represents complex motion trajectories more accurately while still using a single motion vector per block for coding efficiency.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If motion vectors are scaled according to temporal distance between reference and current images, then motion compensation accuracy is improved, but computational complexity increases

Engineering Contradiction:
Improvemotion compensation accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The temporal distance between reference images and current image is pre-calculated, and motion vector scaling factors are determined in advance. This preliminary computation allows the actual motion compensation process to use pre-computed scaling factors, reducing the computational complexity during the main encoding/decoding operation while maintaining accurate motion compensation.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP2553927B1Methods and devices for encoding and decoding an image sequence, which implement prediction by forward motion compensation, and corresponding computer program
Publication Date: 2018.11.28 ORANGE SA
  • EP2553927B1 patent drawingFigure 1~2B
  • EP2553927B1 patent drawingFigure 3
  • EP2553927B1 patent drawingFigure 4A~7

AI summary

The invention relates to a method for encoding an image sequence, said method implementing a step of predicting at least one ordinary image by means of forward motion compensation on the basis of at least one reference image. Said prediction step implements, for at least one ordinary block (32) of said ordinary image, a sub-step (31) of forward projection of at least one block of the reference image onto said ordinary image, thus producing at least one projected block that at least partially covers said ordinary block. According to the invention, the prediction step implements the following sub-steps for said ordinary block (32): compartmentalizing (33) said ordinary block, thus producing at least one sub-block that is at least partially covered by at least one projected block; at least one iteration of the following steps for at least one covered sub-block; verifying (34) that said covered sub-block observes a predetermined assignment criterion; in the event of positive verification, assigning (35) one of said projected movement vectors to said covered sub-block; in the event of negative verification, compartmentalizing (36) said covered sub-block, as long as a predetermined minimum size has not been reached for said covered sub-block.