Volumetric Video Viewpath Forecasting for Distortion-Free Inserts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to adequately insert digital information and advertisements into volumetric video without angular distortion and to remove obstructing objects, ensuring seamless user experience across varying viewpoints.
Innovation Solution
A method that captures multiple fields of view, analyzes excitement levels, forecasts a virtual field of view path, and composites data from cameras to insert images while managing angular offset and object transparency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If digital content is inserted into volumetric video from a fixed camera viewpoint, then the insertion process is simple, but angular distortion occurs when users view from different angles
Solution Approach 1:
The system dynamically adjusts the field of view parameters and camera viewpoints based on the user's viewing angle. The virtual camera's field of view is modified in real-time to compensate for angular offsets, ensuring that inserted digital content remains undistorted regardless of the user's viewing position. This dynamic adaptation resolves the contradiction by making the insertion process angle-independent.
Solution Approach 2:
The system changes key parameters including the virtual camera's field of view angle, focal length, and position to maintain proper perspective for inserted content. By adjusting these parameters based on the user's viewing angle, the system eliminates angular distortion while keeping the insertion process straightforward.
2Adaptability or versatility
If multiple cameras are used to capture volumetric video from different viewpoints, then user viewing flexibility is improved, but system complexity increases
Solution Approach 1:
Instead of using multiple physical cameras to capture all possible viewpoints, the system uses a single physical camera to capture volumetric data and then creates multiple virtual camera viewpoints through computational rendering. This copying approach provides user viewing flexibility while avoiding the complexity of deploying and synchronizing multiple physical cameras.
Solution Approach 2:
The system transitions from a two-dimensional camera plane to a three-dimensional volumetric space, allowing virtual cameras to be positioned anywhere within the captured volume. This dimensional change enables flexible viewpoint selection without adding physical camera complexity.
3Ease of manufacture
If advertisements are placed in fixed locations within the scene, then insertion is straightforward, but visibility is limited for users viewing from different angles
Solution Approach 1:
The system dynamically repositions and re-orientates inserted advertisements based on the user's current viewing angle. As the user moves through different viewpoints, the advertisements automatically adjust their position and orientation within the volumetric scene to remain visible and properly oriented, ensuring consistent visibility while keeping insertion straightforward.
Data Source
AI summary
Video of a plurality of fields of view of a scene is captured, each field of view comprising data of the scene from a different vantage point. An excitement level is determined by analyzing a portion of the captured video. Using the excitement level, a time series of future excitement levels is forecast. Using the time series of future excitement levels, a virtual field of view path of the scene is forecast. An insert image is determined to be included in the virtual field of view path. Captured data from at least two video cameras in the plurality of video cameras is composited into a virtual field of view of the scene. A rendering of the insert image is inserted into the virtual field of view.