3D Face Model Generation via Local and Global Fitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing three-dimensional face reconstruction methods using 3DMM libraries often result in surface models that are not smooth enough or of lower quality, particularly in complex areas like the mouth and nose, due to noise and flaws in the three-dimensional face mesh.
Innovation Solution
A method that aligns a three-dimensional face mesh with a standard face model using face keypoints, performs local fitting to improve the quality of the mesh, and then conducts global fitting to generate a more accurate three-dimensional face model, incorporating non-rigid registration and vertex fitting techniques to enhance the correspondence between the mesh and the model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If fitting is performed based on the 3DMM library using standard face model, then the three-dimensional face model can be generated, but the surface quality is not smooth enough and has lower quality
Solution Approach 1:
The patent divides the fitting process into two distinct stages: local fitting that focuses on keypoint correspondence and deformation transfer, followed by global fitting that performs smooth item fitting across the entire surface. This segmentation allows each stage to optimize for its specific purpose, resulting in both accurate keypoint alignment and smooth surface quality.
Solution Approach 2:
The patent performs preliminary local fitting to establish accurate keypoint correspondence and generate an initial three-dimensional face model before performing the global smooth item fitting. This preliminary action ensures that the subsequent global fitting operates on a well-aligned foundation, improving overall surface quality.
2Measurement precision
If deformation transfer technology is used to pull points from low-model point cloud to high-model point cloud, then correspondence between models is established, but noise and flaws remain in the three-dimensional face mesh
Solution Approach 1:
The patent extracts and focuses on keypoint correspondence through local fitting, separating this critical alignment function from the overall fitting process. By taking out the keypoint matching task and handling it specifically through deformation transfer, the patent ensures accurate correspondence while isolating the noise and flaws to manageable levels.
Solution Approach 2:
The patent applies different fitting qualities to different regions: local fitting with high precision for keypoint correspondence in critical areas, and global smooth fitting for the overall surface. This local quality approach ensures that noise and flaws are minimized in the most important regions while maintaining overall surface smoothness.
3Ease of manufacture
If global fitting is performed directly without local fitting, then the process is simpler, but the three-dimensional face model does not closely match the target object
Solution Approach 1:
The patent segments the fitting process into local and global stages, where local fitting handles keypoint correspondence and global fitting handles overall surface alignment. This segmentation maintains relative simplicity while significantly improving model accuracy compared to direct global fitting.
Solution Approach 2:
The patent performs preliminary local fitting to establish accurate keypoint correspondence before performing global fitting. This preliminary action ensures that the global fitting starts with a well-aligned foundation, improving final model accuracy without adding excessive complexity.
Data Source
AI summary
A three-dimensional face model generation method is provided. The method includes: obtaining an inputted three-dimensional face mesh of a target object; aligning the three-dimensional face mesh with a first three-dimensional face model of a standard object according to face keypoints; performing fitting on the three-dimensional face mesh and a local area of the first three-dimensional face model, to obtain a second three-dimensional face model after local fitting; and performing fitting on the three-dimensional face mesh and a global area of the second three-dimensional face model, to obtain a three-dimensional face model of the target object after global fitting.