Adaptive Motion Vector Prediction for Video Encoding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The motion vector competition method in video coding results in increased code amounts, low encoding efficiency, and large prediction errors, particularly when blocks lack motion vectors or are outside the frame region, as it inefficiently selects candidates from a small region and transmits flags.

Innovation Solution

An image encoding device that performs adaptive motion vector selection by adjusting the number and position of candidate blocks, setting non-referenceable blocks as zero vectors, and generating prediction motion vectors from surrounding blocks, improving encoding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If motion vector competition method is used, then prediction accuracy may be improved, but code amount increases due to flag transmission

Engineering Contradiction:
Improveprediction accuracyVSAvoidcode amount
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts and removes the flag transmission mechanism from the motion vector prediction process. Instead of transmitting flags to indicate prediction mode selection, the system directly transmits the prediction motion vector itself, eliminating unnecessary data transmission while maintaining prediction accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent inverts the conventional approach by not selecting from multiple candidates using flags, but rather directly using the most suitable prediction motion vector without flag-based selection. This reverses the traditional mode-selection paradigm and achieves compression by eliminating the selection indicator overhead.

Inventive Principle:
Principle #13The other way round (Inversion)

2Measurement precision

If motion vector competition method is used, then prediction may be improved, but encoding efficiency decreases due to flag processing

Engineering Contradiction:
Improveprediction qualityVSAvoidencoding efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent removes the flag processing step from the encoding workflow. By eliminating the need to encode and transmit flags for motion vector prediction mode selection, the encoding process becomes more efficient while still achieving accurate motion compensation through direct transmission of prediction vectors.

Inventive Principle:
Principle #2Taking out (Extraction)

3Area of stationary object

If motion vector competition method is used, then local prediction may be improved, but prediction error increases due to limited candidate region

Engineering Contradiction:
Improvecandidate block regionVSAvoidprediction error
Core Design Contradiction:
Area of stationary objectVSMeasurement precision

Solution Approach 1:

The patent expands the candidate block selection from a limited local region to include blocks from multiple reference frames and broader spatial regions. This dimensional expansion in the search space allows the system to find more accurate prediction vectors by considering a wider range of potential candidates beyond the immediate neighboring blocks.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentUS12114007B2Image encoding method, image encoding device, image decoding method, and image decoding device
Publication Date: 2024.10.08 MAXELL LTD
  • US12114007B2 patent drawing
  • US12114007B2 patent drawing
  • US12114007B2 patent drawing

AI summary

For the purpose of providing an image encoding technique for reducing the amount of codes, in inter prediction in which motion search is performed by using a block obtained by dividing an input image and a reference image, the motion vector of the block is predicted by selecting from the motion vectors of the surrounding blocks. Further, in the method of calculating a differential vector and then performing encoding, the number or positions of candidate blocks from which a prediction vector is selected are changed block by block or frame by frame and thereby adaptive and efficient video image compression is performed.