Region-Based Face Tracking via Segmented Rigidity Weights
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing facial tracking systems using linear face shape models struggle with accurate representation of rich or dynamic facial expressions and movements, leading to unstable and jittery head motion due to residual regression errors.
Innovation Solution
The implementation of region-based models that segment different regions of a face and model shape variations separately, incorporating adaptive rigidity weighting and dense motion-guided correctives to improve expressiveness and stability in real-time face tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If linear facial shape models are used for real-time face tracking, then computational efficiency is maintained, but model expressiveness and tracking stability deteriorate due to residual regression errors
Solution Approach 1:
The face is segmented into multiple spatial regions (e.g., forehead, eyes, nose, mouth, chin) and each region is tracked independently with region-specific parameters. This segmentation allows the system to capture local facial expressions and movements more accurately while maintaining overall computational efficiency through parallel processing of regions.
Solution Approach 2:
Different rigidity weights are assigned to different facial regions based on their local properties. Rigid regions (like the forehead and chin) receive higher rigidity weights to maintain stability, while flexible regions (like the mouth and eyes) receive lower weights to allow natural expression variations. This local differentiation improves tracking accuracy without compromising real-time performance.
2Device complexity
If linear facial shape models estimate model parameters together with rigid head poses, then computational complexity is reduced, but measurement precision deteriorates due to coupled estimation errors
Solution Approach 1:
The estimation process is segmented into two distinct phases: first estimating rigid head pose parameters, then estimating regional expression parameters conditioned on the pose estimates. This decoupled estimation reduces algorithmic complexity while improving precision by avoiding the error propagation inherent in fully coupled estimation.
Solution Approach 2:
The system dynamically adjusts the estimation process by first establishing the rigid pose framework and then adding flexible regional expressions. This dynamic two-stage approach allows the system to adapt to different facial configurations and expressions while maintaining computational tractability.
Data Source
AI summary
Aspects of the present disclosure involve a system comprising a computer-readable storage medium storing at least one program and a method for accessing a set of images depicting at least a portion of a face. A set of facial regions of the face is identified, each facial region of the set of facial regions intersecting another facial region with at least one common vertex that is a member of a set of facial vertices. For each facial region of the set of facial regions, a weight formed from a set of region coefficients is generated. Based on the set of facial regions and the weight of each facial region of the set of facial regions, the face is tracked across the set of images.