3D Human Mesh Generation from 2D Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating three-dimensional computer-generated imagery of humans from two-dimensional images are inefficient, as they require manual processing and lack alignment with target application topology, resulting in unsuitable polygonal meshes and textures for interactive video games.
Innovation Solution
A processing workflow that acquires images of a person's head, generates a polygonal mesh and textures, and aligns them with a base mesh using iterative closest point and simulated annealing methods, producing application-resolution models suitable for interactive video games.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual processing methods are used to generate three-dimensional computer-generated imagery from two-dimensional images, then some level of customization and control can be achieved, but the process is inefficient and time-consuming
Solution Approach 1:
The system performs automatic processing where the computer system itself carries out the entire workflow of generating three-dimensional visual objects from two-dimensional images without requiring manual intervention. The system automatically detects anatomical features, generates polygonal meshes, creates textures, and aligns the final model, making the process self-service and highly efficient.
2Manufacturing precision
If traditional processing workflows are used to generate polygonal meshes and textures, then basic three-dimensional models can be created, but they lack alignment with target application topology making them unsuitable for interactive video games
Solution Approach 1:
The patent replaces manual mechanical processing with automated computational methods. Specifically, it uses iterative closest point algorithms and simulated annealing optimization to automatically align the generated polygonal mesh with the target application topology, eliminating the need for manual adjustment while achieving precise alignment suitable for interactive video games.
3Productivity
If automated processing is implemented to improve efficiency, then generation speed increases, but ensuring proper alignment and quality may become more difficult
Solution Approach 1:
The system incorporates feedback mechanisms through iterative optimization algorithms. The iterative closest point method continuously refines the alignment between detected anatomical features and the generated model by comparing and adjusting positions across multiple iterations. Similarly, simulated annealing uses feedback from energy function evaluations to converge on optimal parameter values, ensuring high alignment accuracy while maintaining automated processing speed.
Data Source
AI summary
An example method of generating three-dimensional visual objects representing a person based on two-dimensional images of at least a part of the person's body may include receiving a first polygonal mesh representing a human body part, wherein the first polygonal mesh is compliant with a target application topology. The example method may further include receiving a second polygonal mesh representing the human body part, wherein the second polygonal mesh is derived from a plurality of images of a person. The example method may further include modifying at least one of the first polygonal mesh or the second polygonal mesh to optimize a value of a metric reflecting a difference between the first polygonal mesh and the second polygonal mesh.