3D Model Texture Mapping via Pose-Based Point Correspondence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D face reconstruction technologies face challenges in accurately aligning texture information with 3D models due to deviations in correspondence between 3D points and pixels, especially when images are taken from different angles or when facial movements occur, leading to incorrect texture mapping.
Innovation Solution
The method involves generating multiple 3D networks from different angles, aligning them to a common pose, identifying offsets between corresponding points, and updating the initial correspondence to achieve a more precise alignment between point cloud information and color information, thereby improving texture mapping accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If 3D networks are generated from multiple angles to capture comprehensive texture information, then texture mapping coverage is improved, but correspondence accuracy between 3D points and pixels deteriorates due to alignment deviations
Solution Approach 1:
The patent segments the texture mapping process into multiple independent 3D networks, each generated from a specific angle. Each network maintains its own correspondence relationships, allowing comprehensive coverage while preserving local accuracy. The segmentation enables independent optimization of correspondence in each angular view without compromising overall texture mapping quality.
Solution Approach 2:
The patent introduces a new dimension of angular perspective to the texture mapping process. By generating 3D networks from multiple angles and integrating them through pose information, the system achieves comprehensive texture coverage while maintaining correspondence accuracy through multi-dimensional spatial relationships. The camera pose parameters provide an additional dimensional framework for accurate alignment.
2Manufacturing precision
If 3D networks are aligned to a common pose to improve correspondence, then texture mapping precision is improved, but computational complexity increases due to pose adjustment and offset calculation
Solution Approach 1:
The patent performs preliminary pose adjustment and offset calculation during the 3D network generation phase. By pre-aligning the 3D networks to a common pose framework and calculating necessary offset values beforehand, the system reduces computational complexity during the final texture mapping process. The preliminary actions establish a standardized framework that simplifies subsequent operations.
Solution Approach 2:
The patent introduces camera pose parameters as an intermediary element that mediates between the multiple 3D networks and the final texture mapping process. The pose information serves as a common reference framework that simplifies alignment operations and reduces direct computational complexity. The offset values calculated through pose relationships act as intermediaries that bridge different angular views without requiring complex real-time computations.
3Measurement precision
If correspondence between 3D points and pixels is updated to account for facial movements, then texture alignment accuracy is improved, but processing time increases due to offset calculation and correspondence update
Solution Approach 1:
The patent performs correspondence updates and offset calculations during the 3D network generation phase rather than during final texture mapping. By pre-adjusting correspondences to account for facial movements and pose variations beforehand, the system reduces processing time during the actual texture mapping operation. The preliminary actions establish accurate correspondence relationships that can be directly applied without time-consuming real-time computations.
Solution Approach 2:
The patent creates multiple 3D network copies from different angular views, each with its own correspondence relationships. These copies are then integrated using pose information and offset calculations, allowing the system to maintain accurate correspondence for facial movements without requiring complex real-time adjustments. The copying approach enables parallel processing and reduces overall processing time while maintaining high accuracy.
Data Source
AI summary
A method for acquiring a texture of a three-dimensional (3D) model includes: acquiring at least two 3D networks generated by a target object based on a plurality of angles, the at least two 3D networks including a first correspondence between point cloud information and color information of the target object, and first camera poses of the target object; acquiring an offset between 3D points used for recording the same position of the target object in the at least two 3D networks according to the first camera poses respectively included in the at least two 3D networks; updating the first correspondence according to the offset, to acquire a second correspondence between the point cloud information and the color information of the target object; and acquiring a surface color texture of a 3D model of the target object according to the second correspondence.