Virtual Viewpoint Image Generation for Occluded Region Discrimination

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing techniques for generating virtual viewpoint images from multiple captured images do not accurately account for occluded regions, leading to incomplete representation of blind spots.

Innovation Solution

An information processing apparatus that obtains object information from multiple image capturing apparatuses, discriminates between visible and occluded regions based on this information, and generates virtual viewpoint images that display the results of this discrimination.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If blind spot regions are specified using only viewpoint coordinates and viewing angles, then the calculation is simple, but the occluded regions are not accurately reflected

Engineering Contradiction:
Improvecalculation complexityVSAvoidblind spot region accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent transitions from two-dimensional blind spot specification (viewpoint coordinates and viewing angles) to three-dimensional blind spot specification by incorporating depth information from multiple viewpoint images. This allows accurate representation of occluded regions by considering the spatial position of objects relative to the viewpoint, not just angular relationships.

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

Solution Approach 2:

The patent introduces an intermediary process that uses object recognition results from multiple viewpoint images to determine whether regions are truly occluded. This intermediary analysis bridges the gap between simple geometric blind spot calculation and accurate occlusion detection by considering actual object positions in three-dimensional space.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If multiple viewpoint images are analyzed to accurately detect occluded regions, then blind spot accuracy is improved, but processing complexity increases

Engineering Contradiction:
Improveoccluded region detection accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent performs preliminary object recognition and position determination from multiple viewpoint images before generating the blind spot map. By pre-identifying objects and their three-dimensional positions, the system avoids complex real-time analysis during blind spot calculation, reducing overall processing complexity while maintaining accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent creates a three-dimensional virtual model or representation of the captured scene based on images from multiple viewpoints. This virtual copy allows blind spot analysis to be performed on the modeled scene rather than directly on multiple images, simplifying the processing while preserving accurate occlusion information.

Inventive Principle:
Principle #26Copying

Data Source

PatentUS20250047827A1Information processing apparatus, information processing method, and non-transitory computer-readable storage medium
Publication Date: 2025.02.06 CANON KK
  • US20250047827A1 patent drawing
  • US20250047827A1 patent drawing
  • US20250047827A1 patent drawing

AI summary

An information processing apparatus obtains object information representing a position and a shape of an object in an image capturing region of which image is captured from different directions by a plurality of image capturing apparatuses, discriminates between a visible region that can be seen from a specific position in the image capturing region and an occluded region that cannot be seen from the specific position due to occlusion by the object based on the object information, and generates a virtual viewpoint image that displays information based on a result of the discrimination, the virtual viewpoint image being based on image data based on image capturing by the plurality of image capturing apparatuses and virtual viewpoint information indicating a position and a direction of a virtual viewpoint.