3D Face Texture Consistency via Inner-Outer Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D human scanning technologies often result in artifacts such as misaligned eyes or blurred facial expressions due to averaging texture from multiple frames or inaccurate texture registration, which are noticeable and unappealing.
Innovation Solution
An algorithm that detects a frontal face, maps texture from a single frame to the inner part of the face, and uses depth measurements to ensure consistent facial expressions, with a special scanning scenario requiring the user to start scanning from a frontal position for high-quality images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If texture is synthesized from multiple images, then more complete facial coverage is achieved, but artifacts such as misaligned eyes and blurred expressions occur
Solution Approach 1:
The patent divides the face into two distinct regions: the inner face (eyes, nose, mouth) and the outer face (forehead, cheeks, jawline). This segmentation allows different texturing strategies to be applied to each region, resolving the contradiction by using single-frame texturing for the inner face to maintain precision while using multi-frame synthesis for the outer face to achieve complete coverage.
Solution Approach 2:
The patent applies different quality requirements and processing methods to different parts of the face. The inner face region receives high-precision single-frame texturing to avoid artifacts in critical areas, while the outer face accepts multi-frame synthesis for complete coverage. This local differentiation resolves the contradiction between coverage and precision.
2Manufacturing precision
If texture is mapped from a single frame, then consistent facial expression is achieved, but complete facial coverage cannot be obtained
Solution Approach 1:
The patent segments the face into inner and outer regions, applying single-frame texturing only to the inner face where expression consistency is critical, while using multi-frame synthesis for the outer face to achieve complete coverage. This resolves the contradiction by limiting single-frame mapping to only where it is most needed.
Solution Approach 2:
The patent applies single-frame texturing selectively only to the inner face region rather than the entire face. This partial application achieves expression consistency where it matters most (eyes, nose, mouth) while accepting multi-frame synthesis for the remaining outer regions, balancing precision and coverage.
3Ease of operation
If conventional face detection based on RGB images is used, then face detection is achieved, but accurate frontal face detection for scanning is not guaranteed
Solution Approach 1:
The patent uses depth information as an intermediary to enhance face detection accuracy. By incorporating depth measurements from the 3D scanner, the system can more accurately determine whether a face is frontal and suitable for scanning, resolving the contradiction between simple detection and accurate frontal face identification.
Data Source
AI summary
The present invention provides a method for the 3D scanning of a person's head.