A method and device for 3D facial modeling in medical aesthetics

By combining multi-view image processing and joint optimization with parametric meshes and 3D Gaussian representation, a 3D face model with both measurable meshes and high-fidelity appearance is generated. This solves the accuracy and stability problems of 3D face reconstruction in medical aesthetics scenarios in existing technologies, and achieves efficient medical aesthetics measurement and realistic visual effects.

CN123118965APending Publication Date: 2026-11-17SOUTH CHINA UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611334078.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-31
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing 3D face reconstruction technology has difficulty accurately restoring the profile, nose, jawline, preauricular contour, and subtle skin undulations in medical aesthetic scenarios. Furthermore, pure Gaussian representation lacks unified topology and medical aesthetic measurement capabilities, making it difficult to generate 3D face models that combine realistic visual effects with measurement value.

Method used

By processing images from multiple perspectives, a unified standard space is established, identity feature vectors and view-level parameters are integrated, a joint optimization objective loss function is constructed, and a 3D face model with both measurable mesh and high-fidelity appearance is generated by combining parameterized mesh and 3D Gaussian representation. Confidence and source label auditing are then performed.

Benefits of technology

It achieves cross-view uniqueness, spatial measurement capabilities, and photorealistic appearance of 3D face models, ensuring that AI-completed areas are not misused for medical measurements, improving the geometric stability of the side face, jawline, nose, and midface, reducing color difference and occlusion misprojection, and providing traceable skin texture.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118965A_ABST
    Figure CN123118965A_ABST
Patent Text Reader

Abstract

This invention relates to the field of medical aesthetic modeling technology, specifically a method and device for 3D facial modeling in medical aesthetics. The method includes: acquiring original facial images from multiple perspectives and obtaining camera parameters; establishing initial 3D facial values ​​within a unified standard space; extracting identity feature vectors from each view and weighted fusing them to obtain initial values ​​for shared identity shape parameters; constructing a joint optimization objective loss function to jointly optimize the shared identity shape parameters and view-level parameters; enhancing the details of effective facial regions through depth and 3D Gaussian enhancement; back-projecting and fusing the original high-definition images from multiple perspectives to generate UV texture maps; performing real-time confidence measurement and source auditing based on the optimized shared identity parameters and independent expression parameters, pose parameters, and UV texture maps for each view; and outputting a standardized 3D facial model and accompanying metadata. This invention can balance measurable meshes and high-fidelity appearance, and prevents generated content from being mistakenly used as real measurement data through confidence levels and source tags.
Need to check novelty before this filing date? Find Prior Art