Virtual Object Fitting for Human Faces Using Facial Feature Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods for adding 3D human head ornaments fail to accurately fit virtual objects to individual human faces, neglecting facial particularities such as fatness and contour shape, resulting in poor visual effects.
Innovation Solution
A method and device that acquire a virtual object with a corresponding standard model, determine control information based on the target object, and generate a second virtual object matched with the target object, ensuring better fitting by considering the specific characteristics of the face, such as bone points and contour features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a standard model is used for virtual objects, then the production process is simplified and efficiency is improved, but the fitting accuracy to individual faces deteriorates
Solution Approach 1:
The patent applies local quality by extracting specific facial features (bone points, contour points) from individual faces and applying localized transformations to the virtual object. Instead of using a uniform standard model for all faces, the system adjusts specific regions of the virtual object based on individual facial characteristics, thereby maintaining production efficiency while improving fitting accuracy to individual faces.
2Manufacturing precision
If face recognition and key point detection are performed, then the fitting accuracy is improved, but the processing time and complexity increase
Solution Approach 1:
The patent applies preliminary action by pre-establishing the correspondence relationships between standard model points and facial feature points. The control information template is prepared in advance, containing the mapping relationships between virtual object points and face points. During actual processing, this pre-prepared template is directly applied, significantly reducing the processing time while maintaining high fitting accuracy.
3Device complexity
If rigid transformation matrix is used for fitting, then the processing complexity is reduced, but the visual effect deteriorates due to inability to capture facial particularities
Solution Approach 1:
The patent applies dynamics by transitioning from a static rigid transformation matrix to a dynamic adaptive transformation process. The system uses control information derived from actual facial features to dynamically adjust the virtual object's transformation parameters. This allows the fitting process to adapt to different facial characteristics while maintaining reasonable processing complexity through automated control information generation.
Data Source
AI summary
Embodiments of the present disclosure relates to a method and device for processing an image, and a nonvolatile storage medium. The method includes: acquiring a first image and a first virtual object having a corresponding relationship with a preset standard model, the first image includes a target object; determining control information based on the target object and the standard model; obtaining a second virtual object matched with the target object by processing the first virtual object based on the control information; and generating a second image; the second image includes the target object fitted with the second virtual object.