Motion Compensation Prediction for Spherical Panoramic Images

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Motion-compensated prediction in spherical panoramic images faces challenges due to geometric deformation when reference pixels are not on the same projection plane, leading to inaccurate pixel value prediction across polyhedral formats.

Innovation Solution

The method determines a target reference pixel by finding a spherical reference location corresponding to the initial reference pixel, which is then used to calculate a polyhedral reference location, allowing for accurate prediction of pixel values across different sub-images in a polyhedral format.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If motion-compensated prediction is performed using reference pixels from different sub-images in polyhedral format, then the coverage of reference pixels is improved, but the prediction accuracy deteriorates due to geometric deformation at face boundaries

Engineering Contradiction:
Improvereference pixel coverageVSAvoidprediction accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent introduces a spherical coordinate system as an intermediary to map reference pixel locations. By converting polyhedral face coordinates to spherical coordinates and back, the system accurately determines corresponding reference pixels across face boundaries while accounting for geometric deformations, thus maintaining prediction accuracy while expanding reference pixel coverage

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent changes the coordinate parameter system from simple polyhedral face coordinates to spherical coordinates. This parameter transformation allows accurate representation of reference pixel locations across different faces by considering the spherical geometry, thereby resolving the prediction accuracy issue when using reference pixels from multiple sub-images

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If reference pixels are selected from adjacent faces to expand search range, then motion estimation flexibility is improved, but prediction error increases due to boundary deformation

Engineering Contradiction:
Improvemotion estimation flexibilityVSAvoidprediction reliability
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The spherical coordinate system acts as a mediator to accurately map reference pixel positions across face boundaries. By using spherical coordinates to represent directions and converting back to polyhedral coordinates, the system maintains reliable pixel correspondence even when reference pixels are selected from adjacent faces, thus preserving prediction reliability while improving motion estimation flexibility

Inventive Principle:
Principle #24Intermediary (Mediator)

3Device complexity

If simple planar projection is used for polyhedral images, then processing complexity is reduced, but coding efficiency deteriorates due to boundary inconsistency

Engineering Contradiction:
Improveprocessing complexityVSAvoidcoding efficiency
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent changes the coordinate parameter representation from simple planar coordinates to spherical coordinates. This parameter change enables accurate handling of boundary pixels across different faces, improving coding efficiency by reducing prediction errors at boundaries while maintaining manageable processing complexity through systematic coordinate transformation

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3515076B1Motion compensation prediction method and device
Publication Date: 2021.11.10 HUAWEI TECH CO LTD
  • EP3515076B1 patent drawingFigure 1~2(k)
  • EP3515076B1 patent drawingFigure 3
  • EP3515076B1 patent drawingFigure 4(a)~4(b)

AI summary

Embodiments of this application provide a motion-compensated prediction method and a motion-compensated prediction apparatus. The motion-compensated prediction method includes: determining a location that is of an initial reference pixel of a current pixel and that is in a reference image, where the current pixel is located in a first sub-image in the current image; when the initial reference pixel is located outside a second sub-image that is in the reference image and that is at a location corresponding to the first sub-image, determining, based on the location of the initial reference pixel, a location that is of a target reference pixel of the current pixel and that is in the reference image; and determining a prediction value of a pixel value of the current pixel based on a pixel value of the target reference pixel and/or a pixel value of a neighboring pixel of the target reference pixel. In the embodiments of this application, accuracy of motion-compensated prediction can be improved.