Autostereo Tapestry Encoding for Multi-View Image Reconstruction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current autostereoscopic image processing methods face challenges in efficiently transmitting and reconstructing multiple views of a scene, particularly in handling occluded regions with complex textures and gradients, which leads to artifacts and reduced image quality.

Innovation Solution

The proposed method generates an autostereo tapestry image by combining leftmost and rightmost view images with disparity and occlusion maps, creating displacement maps to encode and decode multiple views, thereby reducing the need for pixel infilling and minimizing artifacts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of energy

If conventional autostereoscopic image processing methods are used to transmit and reconstruct multiple views, then bandwidth efficiency is improved, but image quality deteriorates due to artifacts in occluded regions

Engineering Contradiction:
Improvebandwidth efficiencyVSAvoidimage quality
Core Design Contradiction:
Loss of energyVSManufacturing precision

Solution Approach 1:

The patent segments the scene into different view components by generating separate displacement maps for leftmost and rightmost views, and creating an occlusion map to identify occluded regions. This segmentation allows independent processing of different view components, enabling selective handling of occluded regions to maintain image quality while managing bandwidth efficiently.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces displacement maps and occlusion maps as intermediary data structures that facilitate the reconstruction of multiple views from a single encoded image. These intermediaries enable accurate pixel mapping and occlusion handling, improving image quality without requiring transmission of multiple full-resolution images, thus maintaining bandwidth efficiency.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Manufacturing precision

If multiple full-resolution view images are transmitted to ensure image quality, then image quality is improved, but bandwidth consumption increases

Engineering Contradiction:
Improveimage qualityVSAvoidbandwidth consumption
Core Design Contradiction:
Manufacturing precisionVSQuantity of substance

Solution Approach 1:

Instead of transmitting multiple full-resolution images, the patent transmits a single encoded image that contains compressed representation of multiple views. The displacement maps and occlusion maps serve as compact copies that enable reconstruction of full-resolution views at the decoder, significantly reducing bandwidth consumption while maintaining image quality.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent changes the representation parameters by encoding spatial relationships through displacement maps rather than transmitting complete image data. This parameter transformation allows the system to convey essential view information in a compressed form, reducing bandwidth consumption while enabling high-quality reconstruction through the encoded parameters.

Inventive Principle:
Principle #35Parameter changes

3Loss of information

If pixel infilling is used to reconstruct occluded regions, then completeness of view is improved, but artifacts increase and image quality deteriorates

Engineering Contradiction:
Improvecompleteness of viewVSAvoidimage quality
Core Design Contradiction:
Loss of informationVSManufacturing precision

Solution Approach 1:

The patent performs preliminary identification of occluded regions through occlusion maps before reconstruction. By pre-identifying which regions are occluded in each view, the system can selectively process these regions using appropriate algorithms, ensuring completeness while minimizing artifacts through targeted rather than blanket infilling approaches.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses occlusion maps as feedback information to guide the reconstruction process. The occlusion maps provide real-time information about which regions require special handling, allowing the reconstruction algorithm to adapt its behavior and apply appropriate techniques only where needed, thus maintaining image quality while ensuring completeness of occluded regions.

Inventive Principle:
Principle #23Feedback

4Loss of energy

If a single view image is transmitted to reduce bandwidth, then bandwidth efficiency is improved, but flexibility in rendering different viewpoints is reduced

Engineering Contradiction:
Improvebandwidth efficiencyVSAvoidflexibility in viewpoint rendering
Core Design Contradiction:
Loss of energyVSAdaptability or versatility

Solution Approach 1:

The patent creates a universal encoded image representation that can serve multiple viewing angles simultaneously. The displacement maps encode spatial relationships that enable the single image to be reconstructed into multiple different viewpoints, providing multi-functionality that maintains bandwidth efficiency while restoring flexibility for rendering various perspectives.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent adds dimensional information through displacement maps that encode depth and spatial relationships. This additional dimension allows the system to reconstruct multiple viewpoints from a single 2D image, effectively adding a third dimension of information that enables flexible viewpoint rendering while maintaining bandwidth efficiency of single-image transmission.

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

Data Source

PatentEP2822279B1Autostereo tapestry representation
Publication Date: 2019.10.16 DOLBY LABORATORIES LICENSING CORP
  • EP2822279B1 patent drawingFigure 1
  • EP2822279B1 patent drawingFigure 2
  • EP2822279B1 patent drawingFigure 3

AI summary

Representation and coding of multi-view images using tapestry encoding are described. A tapestry comprises information on a tapestry image, a left-shift displacement map and a right-shift displacement map. Perspective images of a scene can be generated from the tapestry and the displacement maps. The tapestry image is generated from a leftmost view image, a rightmost view image, a disparity map and an occlusion map.