Light Field Occlusion Detection via Depth Map Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Light field capturing systems face challenges in detecting occlusions in real-time due to the reliance on positional dependencies during image capture, which limits the ability to render background objects hidden by foreground objects, and existing methods are computationally intensive, making real-time occlusion detection difficult.
Innovation Solution
A method that calculates a depth map and visibility map for pixels in a reference image, determines candidate areas for potential occlusions based on depth map segmentation, and assesses occlusion states using visibility values from neighborhoods, allowing for real-time detection and user notification during image capture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional occlusion detection methods are used, then occlusion detection capability is provided, but computational load is too high for real-time processing
Solution Approach 1:
The patent segments the image into multiple depth intervals and processes each interval separately to identify occlusions. By dividing the image into foreground and background depth intervals, the method reduces computational complexity while maintaining occlusion detection accuracy, enabling real-time processing of light field images.
Solution Approach 2:
The patent performs preliminary depth map calculation and segmentation before occlusion detection. By pre-processing the image to establish depth intervals and identify candidate occlusion areas in advance, the method reduces the computational burden during actual occlusion detection, facilitating real-time application.
2Measurement precision
If light field capturing system is used, then multiple viewpoints are available for occlusion detection, but device complexity increases
Solution Approach 1:
The patent uses a single light field capturing system that can capture images from multiple viewpoints simultaneously. By making the capturing system multi-functional (able to capture multiple views in one operation), the method achieves improved occlusion detection accuracy without proportionally increasing device complexity, as the same hardware serves multiple detection purposes.
3Measurement precision
If depth map segmentation is performed to identify candidate areas, then occlusion detection precision is improved, but processing time increases
Solution Approach 1:
The patent applies different processing strategies to different regions of the image based on depth interval segmentation. By focusing detailed occlusion detection only in candidate areas identified through depth segmentation, and using simplified methods for other regions, the method maintains high detection precision while reducing overall processing time for real-time application.
Data Source
AI summary
A method is proposed for detecting occlusions in an image captured by a light field capturing system, comprising, for at least one reference image belonging to a matrix of images captured by the light field capturing system: —calculating a depth map and a visibility map for pixels in the reference image; —determining at least one candidate area in the reference image in which the potential occlusion may occur based at least on a segmentation of the depth map; —determining an information representative of an occlusion state in the at least one candidate area based at least on visibility values of the visibility map associated with at least two neighborhoods of the at least one candidate area.