Image-Fitted 3D Face Models for Large-Angle Pose Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face pose estimation methods using three-dimensional average face models suffer from reduced accuracy when dealing with large-angle and exaggerated facial expressions.
Innovation Solution
Construct a three-dimensional face model corresponding to a two-dimensional face image using a face shape and/or facial expression fitting algorithm, incorporating three-dimensional face prior data to align and fit feature points, thereby improving robustness and accuracy of pose estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a three-dimensional average face model is used for pose estimation, then the method is simple and computationally efficient, but the accuracy deteriorates for large-angle and exaggerated facial expressions
Solution Approach 1:
The patent applies dynamics by transitioning from a static three-dimensional average face model to a dynamic three-dimensional face model that can adapt to various facial expressions and poses. The system reconstructs a three-dimensional face model from a two-dimensional face image, allowing the model to dynamically adjust to different facial configurations rather than relying on a fixed average model, thereby maintaining high accuracy across diverse facial conditions.
Solution Approach 2:
The patent utilizes parameter changes by transforming the face representation from a fixed three-dimensional average model to a parameterized three-dimensional model that can be adjusted based on the specific two-dimensional image input. By extracting feature points and reconstructing the face model with adjustable parameters, the system can adapt to different facial geometries and expressions, resolving the contradiction between model simplicity and estimation accuracy.
2Loss of time
If a three-dimensional average face model is used, then the computation time is short, but the robustness deteriorates for diverse facial expressions and poses
Solution Approach 1:
The patent applies preliminary action by pre-establishing a three-dimensional face model framework that can be quickly instantiated and adjusted for different inputs. The system pre-processes the two-dimensional image to extract key feature points and then efficiently reconstructs the corresponding three-dimensional model using the pre-defined framework, reducing computation time while maintaining robustness for diverse facial expressions and poses.
Solution Approach 2:
The patent utilizes copying by creating a three-dimensional copy of the face from the two-dimensional image input. Instead of using a fixed average model, the system generates a personalized three-dimensional representation that mirrors the specific facial geometry in the input image, enabling accurate pose estimation for diverse expressions while maintaining computational efficiency through optimized reconstruction algorithms.
3Measurement precision
If the three-dimensional model is constructed to match the two-dimensional image accurately, then the pose estimation accuracy improves, but the device complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the face model construction process into distinct segments: extracting feature points from the two-dimensional image, reconstructing the three-dimensional face model based on these points, and determining pose parameters. This segmentation allows each step to be optimized independently, maintaining high accuracy while managing complexity through modular processing stages.
Solution Approach 2:
The patent utilizes an intermediary approach by introducing a three-dimensional face model as an intermediate representation between the two-dimensional image and the final pose estimation. This intermediary model serves as a bridge that translates two-dimensional image data into three-dimensional spatial understanding, enabling accurate pose estimation without directly complexifying the overall system architecture.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This application discloses methods, apparatus, and electronic devices for face pose estimation and three-dimensional face reconstruction. The face pose estimation method comprises: acquiring a two-dimensional face image for processing, constructing a three-dimensional face model corresponding to the two-dimensional face image, and determining a face pose of the two-dimensional face image based on face feature points of the three-dimensional face model and face feature points of the two-dimensional face image. With this approach, the face pose estimation is performed based on the three-dimensional face model corresponding to the two-dimensional face image, instead of only based on a three-dimensional average face model. As a result, a high accuracy pose estimation can be obtained even for a face with large-angle and exaggerated facial expressions. Thus, robustness of the pose estimation can be effectively improved.