Camera Selection for Virtual View Synthesis Without Geometry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for synthesizing virtual view images using actual image-based images face challenges in providing 6 degrees of freedom, leading to high computational complexity and memory usage, especially when using omnidirectional light field images without geometry information.
Innovation Solution
A method and apparatus for selecting cameras that determine whether they are within the field of view at a virtual view position by comparing the direction from the virtual view position to the camera's position, allowing for the selection of suitable input images for virtual view synthesis without relying on geometric information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If omnidirectional light field images are used for virtual view synthesis, then photo-realistic virtual view images can be provided, but computational complexity and memory usage significantly increase
Solution Approach 1:
The patent divides the omnidirectional light field image into multiple view images captured by separate cameras positioned at different locations. Instead of processing the entire omnidirectional image, the system segments it into manageable view images that can be individually selected and processed, significantly reducing computational complexity while maintaining image quality.
Solution Approach 2:
The patent extracts only the necessary view images from the omnidirectional light field based on the virtual camera's field of view. By taking out only the relevant portions (view images within FoV) rather than processing the complete omnidirectional image, the system reduces memory usage and computational load while preserving photo-realistic quality.
2Reliability
If all cameras are considered for virtual view synthesis, then complete coverage is achieved, but processing time and computational resources increase
Solution Approach 1:
The patent applies local quality by selectively processing only those view images that fall within the virtual camera's field of view, rather than uniformly processing all captured images. This localized approach maintains coverage completeness for the relevant viewing area while improving processing efficiency by excluding unnecessary images.
Solution Approach 2:
The patent uses partial action by selecting only the subset of view images necessary for the current virtual view synthesis, rather than using all available images. This partial selection based on FoV criteria achieves sufficient coverage for the virtual viewpoint while significantly reducing processing time and resource consumption.
Data Source
AI summary
The present disclosure provides an apparatus and a method for selecting camera providing input images to synthesize virtual view images. According to the present disclosure, A method of selecting a camera providing as input image to synthesize a virtual view image, the method may comprise, for a camera providing an input image, determining whether or not the camera is comprised in a field of view (FoV) at a virtual view position and in response to the camera determined to be comprised in the field of view, selecting the camera to synthesize the virtual image, wherein the determining determines, by way of comparison, whether or not a direction from the virtual view position to a position of the camera is in the FoV at the virtual view position.