Point Cloud Atlas Encoding With Rotated Low-Distortion Projections
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in encoding point clouds for volumetric video, particularly in minimizing distortion and pixel usage while maintaining high-quality rendering for immersive experiences, especially in 6DoF video where head translations are allowed.
Innovation Solution
The method involves projecting a part of the 3D scene within a solid angle onto low-distortion areas of a projection map, generating depth and color images, and encoding them with associated data indicating rotation, using equirectangular projection mapping and permutation of axis order to minimize distortion and optimize pixel usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If points of the 3D scene are projected onto a projection map using conventional mapping methods, then the entire 3D scene can be represented, but distortion increases and pixel usage becomes inefficient
Solution Approach 1:
The patent divides the 3D scene into multiple point clouds, each representing a specific solid angle. Each point cloud is projected onto a dedicated low-distortion area of the projection map. This segmentation allows each region to be rendered with minimal distortion while efficiently utilizing the projection map space.
Solution Approach 2:
The patent applies different projection strategies to different regions of the 3D scene based on their specific characteristics. Each point cloud is projected onto areas optimized for low distortion, ensuring that each local region maintains high rendering quality appropriate to its spatial characteristics.
2Quantity of substance
If conventional projection methods are used to encode point clouds, then all points can be represented, but pixel requirements increase and encoding efficiency decreases
Solution Approach 1:
The patent segments the point cloud data into multiple point clouds corresponding to different solid angles. Each segmented point cloud is projected onto a specific low-distortion area, which reduces the total pixel requirements compared to projecting the entire point cloud uniformly.
Solution Approach 2:
The patent transforms the projection approach by mapping point clouds from three-dimensional space onto two-dimensional projection map areas that are specifically designed to minimize distortion. This dimensional transformation with optimized mapping reduces the pixel density required while maintaining complete point cloud coverage.
3Manufacturing precision
If the solid angle is rotated to optimize projection, then distortion is minimized, but the complexity of tracking and processing increases
Solution Approach 1:
The patent performs preliminary rotation of the solid angle to an optimal orientation before projection. By pre-orienting the solid angle to minimize distortion, the system avoids complex real-time adjustments during rendering, thereby reducing processing complexity while maintaining high projection accuracy.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
A colored 3D scene is encoded as one or two patch atlas images. Points of the 3D scene belonging a part of the space defined according to a truncated sphere center on a point of view and visible from this point of view are iteratively projected onto projection maps. At each iteration the projected part is removed from the 3D scene and the truncated sphere defining the next part of the scene to be projected is rotated. Once the entirety of the 3D scene is projected on a set of projection maps, pictures are determined within these maps. A picture, also called patch, is a cluster of depth consistent connected pixels. Patches are packed in a depth and a color atlas associated with data comprising an information relative to the rotation of the truncated sphere, so, a decoder can retrieve the projection mapping and proceed to the inverse projection.