Stereoscopic 3D Rendering via Sphere-Based Vertex Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for rendering 360° stereoscopic VR video are inefficient, leading to processor-intensive and memory-intensive processes, with existing techniques failing to seamlessly render 3D objects and causing stereoscopic parallax issues near the zenith and nadir, resulting in user discomfort.
Innovation Solution
The method involves tessellating 3D objects to identify vertices and transforming their positions based on interaxial and convergence parameters, using a projection sphere for equirectangular representations, allowing for single-pass rendering without stitching, thus reducing computational load and avoiding parallax at the poles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If six 90°×90° views are stitched together with offsetting to create artificial interaxial separation, then stereoscopic depth perception is achieved, but the rendering process becomes processor intensive and the results are not seamless
Solution Approach 1:
The patent divides the 360° scene into six 90°×90° views (front, back, left, right, top, bottom) and processes each segment independently. This segmentation allows parallel rendering of multiple views simultaneously, improving productivity while maintaining the stereoscopic effect through controlled offsetting of specific views.
Solution Approach 2:
The patent combines the six rendered views into a complete equirectangular projection by stitching them together. The merging process integrates the segmented views while applying distortion corrections at stitch lines, achieving both efficient processing through segmentation and seamless results through careful merging.
2Manufacturing precision
If a slit scan method is used to render 360° stereoscopic views by repeating 180° high by 1° views for each Y-axis rotation, then complete equirectangular projections are generated, but the process becomes very slow and memory intensive
Solution Approach 1:
Instead of rendering 360 views at 1° resolution, the patent segments the scene into six 90°×90° views. This reduces the total number of renderings from 360 to 6, dramatically reducing rendering time and memory requirements while still achieving complete equirectangular coverage through strategic stitching.
Solution Approach 2:
The patent renders only the necessary six orthogonal views rather than all 360 possible 1° rotations. This partial action approach achieves sufficient coverage for equirectangular projection without the excessive computational cost of rendering every possible view angle at high resolution.
3Reliability
If the zenith and nadir contain stereoscopic parallax, then 3D depth perception is maintained, but vertical disparity causes eye strain, pain, or potentially nausea
Solution Approach 1:
The patent applies different rendering quality to different regions of the equirectangular projection. Specifically, the zenith and nadir regions are rendered without stereoscopic parallax (monoscopic), while the equatorial regions maintain stereoscopic depth perception. This local differentiation eliminates vertical disparity discomfort at the poles while preserving useful depth perception at the equator.
Solution Approach 2:
Instead of applying uniform stereoscopic parallax throughout the entire 360° view, the patent inverts the approach by removing parallax from specific problematic regions (zenith and nadir) while maintaining it in other regions. This selective inversion resolves the conflict between depth perception and user comfort.
Data Source
AI summary
In one implementation, a method involves tessellating a surface of a 3D object by identifying vertices having 3D positions. The method transforms the 3D positions into positions for a first sphere-based projection for a left eye viewpoint and positions for a second sphere-based projection for a right eye viewpoint. Transforming the 3D positions of the vertices involves transforming the vertices based on a user orientation (i.e., camera position) and differences left and right eye viewpoints (e.g., based on interaxial distance and convergence angle). The method further renders a stereoscopic 360° rendering of the 3D object based on the first sphere-based projection for the left eye viewpoint and the second sphere-based projection for the right eye viewpoint. For example, an equirectangular representation of the first sphere-based projection can be combined with an equirectangular representation of the second sphere-based projection to provide a file defining a stereoscopic 360° image.