Face Image Processing via Geometric Key-Point Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face image processing methods using neural networks often result in undesired transformations and degraded effects when attempting to change the age of a face, leading to uncontrollable factors and poor robustness in the transformation process.
Innovation Solution
A method involving the acquisition of first-key-point information, position transformation to obtain second-key-point information conforming to a different facial geometric attribute, and subsequent facial texture coding using a neural network to achieve a more controlled and robust face image transformation, separating geometric and texture attribute transformations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If neural networks are used to transform face images, then transformation capability is achieved, but transformation control and robustness deteriorate
Solution Approach 1:
The patent segments the face transformation process into two independent parts: geometric attribute transformation (using key-point information and transformation matrices) and texture attribute transformation (using neural networks). This segmentation allows each part to be controlled separately, improving overall transformation control and robustness while maintaining the versatility of neural network-based transformation.
2Manufacturing precision
If neural networks process face images, then transformation effect is achieved, but randomness and uncontrollable factors increase
Solution Approach 1:
The patent performs preliminary geometric transformation on key-point information before texture transformation. By pre-establishing the geometric structure through transformation matrices and key-point mapping, the neural network only needs to handle texture attributes, which reduces its randomness and improves the stability and controllability of the overall transformation process.
3Adaptability or versatility
If geometric and texture transformations are combined, then comprehensive transformation is achieved, but control precision deteriorates
Solution Approach 1:
The patent separates geometric attribute transformation and texture attribute transformation into distinct processing stages. Geometric transformation is controlled through key-point information and transformation matrices with high precision, while texture transformation is handled by neural networks. This segmentation maintains control precision for geometric attributes while achieving comprehensive transformation effects.
Data Source
AI summary
Provided are a face image processing method and apparatus, an image device, and a storage medium. The face image processing method includes: acquiring first-key-point information of a first face image; performing position transformation on the first-key-point information to obtain second-key-point information conforming to a second facial geometric attribute, the second facial geometric attribute being different from a first facial geometric attribute corresponding to the first-key-point information; and performing facial texture coding processing by utilizing a neural network and the second-key-point information to obtain a second face image.