Template-Matched Video Coding Without Motion Vectors

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video encoding methods, such as H.264, require motion vectors for inter-frame prediction, leading to increased information volume and complexity, necessitating a more efficient compression technique.

Innovation Solution

A video encoding device and method that divides frames into blocks, searches for highly correlated pixel groups adjacent to each block, and encodes the difference between the block and its prediction, using template matching to determine prediction signals without motion vectors, and adjusts block sizes based on spatial continuity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If motion vectors are used for inter-frame prediction, then prediction accuracy is improved, but information volume increases and encoding complexity increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidinformation volume
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts and removes motion vectors from the encoding system, replacing them with template matching. By taking out the motion vector component entirely, the patent eliminates the associated information overhead while maintaining prediction functionality through direct template comparison with reconstructed blocks.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent uses copying by creating templates from reconstructed blocks and searching for matching templates in reference frames. Instead of encoding motion vectors, the system copies template patterns and uses their spatial-temporal relationships to generate prediction blocks, thereby reducing information volume while preserving prediction accuracy.

Inventive Principle:
Principle #26Copying

2Measurement precision

If motion vectors are used for inter-frame prediction, then prediction accuracy is improved, but device complexity increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidencoding complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent removes motion vectors from the encoding pipeline, simplifying the device architecture. By extracting this component, the patent eliminates the need for motion vector calculation, encoding, and decoding modules, thereby reducing device complexity while maintaining prediction functionality through template matching.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent replaces complex motion vector processing with simpler template copying operations. The system copies reconstructed blocks to form templates and searches for matching templates, which is computationally simpler than traditional motion estimation and compensation, thus reducing encoding complexity.

Inventive Principle:
Principle #26Copying

3Measurement precision

If template size is increased to improve prediction accuracy, then prediction quality improves, but computational load increases

Engineering Contradiction:
Improveprediction qualityVSAvoidcomputational load
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

The patent segments the template matching process by dividing the search into manageable steps: generating templates from reconstructed blocks, searching for matching templates in reference frames, and using the matched template positions to create prediction blocks. This segmentation allows for efficient computation while maintaining high prediction quality.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary actions by pre-generating templates from reconstructed blocks before the actual prediction process. These templates are stored and reused during encoding, which reduces computational load during the prediction phase while maintaining high prediction accuracy through the use of pre-processed template information.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP2627091B1Video coding using template matching
Publication Date: 2016.09.07 NTT DOCOMO INC
  • EP2627091B1 patent drawingFigure 1
  • EP2627091B1 patent drawingFigure 2
  • EP2627091B1 patent drawingFigure 3

AI summary

The present invention enables an efficient encoding of video data in which a prediction signal can be determined without using a motion vector. A video encoding device 100 comprises a region division section 101 for dividing a frame image constituting video data into a plurality of regions as encoding target regions, an encoding section 104 for encoding an image of each region, an inverse transformation section 105 and an addition section 106 for generating reproduced image of the encoded image, a storage section 107 for storing reproduced images, a prediction generation section 108 for searching a region which is highly correlated to a reproduced image of a template region, which is adjacent to the region of the encoding target image in a predetermined positional relationship and is a part of the reproduced image, from the reproduced image, and determining a prediction signal based on the searched region and the above-mentioned positional relationship, and a subtraction section 102 for generating a difference signal between the prediction signal and the encoding target image as a signal for encoding.