Image Encoding Using Motion-Aware Visual Distortion Metrics

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video encoding methods struggle to effectively account for movement across images in a video stream, leading to suboptimal visual quality and increased bitrate due to limitations in distortion metrics and computational complexity.

Innovation Solution

A novel image encoding method that uses multiple motion estimation vectors and prediction vectors to refine visual distortion measurement, allowing for better consideration of movement across images, and integrates a visual distortion metric into the bitrate-distortion decision criterion to optimize encoding modes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple motion estimation vectors and prediction vectors are used to refine visual distortion measurement, then visual quality is improved, but computational complexity increases

Engineering Contradiction:
Improvevisual distortion measurement accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the motion estimation process into multiple vectors (first motion estimation vector, second motion estimation vector, etc.) corresponding to different reference images. Each vector independently estimates motion to a specific reference image, allowing parallel computation and more comprehensive motion characterization without exponentially increasing complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs motion estimation to multiple reference images in advance, generating multiple motion estimation vectors before the actual encoding decision. This preliminary action allows the visual distortion metric to be accurately computed using pre-available vectors, avoiding computationally expensive operations during the rate-distortion optimization phase.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If visual distortion metric is integrated into bitrate-distortion decision criterion, then visual quality is improved, but encoding complexity increases

Engineering Contradiction:
Improvevisual quality assessmentVSAvoidencoding complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent modifies the traditional rate-distortion criterion by incorporating a visual distortion metric that uses motion estimation vectors and prediction vectors. This parameter change transforms the decision criterion from purely pixel-based comparison to a motion-aware metric, improving visual quality assessment while maintaining compatibility with existing encoding frameworks.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If motion compensation is performed using multiple reference images, then accuracy in accounting for movement is improved, but bitrate increases

Engineering Contradiction:
Improvemovement accounting accuracyVSAvoidbitrate
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent computes motion estimation vectors to multiple reference images (excessive action) but only transmits the necessary information for the selected reference image during encoding. The additional vectors are used internally for accurate visual distortion measurement and mode selection, not for transmission, thus improving accuracy without proportionally increasing bitrate.

Inventive Principle:
Principle #16Partial or excessive action

4Quantity of substance

If objective distortion metrics are used for optimization, then encoder output bitrate is reduced, but visual rendering quality deteriorates

Engineering Contradiction:
Improveencoder output bitrateVSAvoidvisual rendering quality
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent introduces a feedback mechanism where the visual distortion metric, computed using motion estimation vectors and prediction vectors, guides the encoding mode selection. This feedback loop ensures that encoding decisions are made based on motion-aware quality assessment rather than purely on bitrate reduction, preventing visual quality deterioration even when using aggressive compression.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3225029B1Image encoding method and equipment for implementing the method
Publication Date: 2023.09.13 ATEME
  • EP3225029B1 patent drawingFigure 1
  • EP3225029B1 patent drawingFigure 2a~2b
  • EP3225029B1 patent drawingFigure 2c

AI summary

The invention proposes a method for encoding a first image in a set of images, in which the first image is cut into blocks, each block being encoded according to one from a plurality of encoding modes comprising at least one temporal correlation prediction encoding mode using a plurality of images from the set of images. The method comprises, for a current block of the first image: determining a first motion estimation vector of the current block, determining a value of a visual distortion metric for a temporal correlation prediction encoding mode, the visual distortion metric using a distance between a first temporal prediction vector of the current block pointing to the second image and the first motion estimation vector of the current block, and using the value of the visual distortion metric to encode the current block.