一种基于标准先验的牙齿图像修复方法和系统
By improving the Auto-Encoder structure and using a standard prior encoder, combined with facial landmark detection and multi-loss function training, the problem of insufficient data in dental image restoration is solved, achieving high-quality dental image restoration results while reducing training costs and data requirements.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHENDU PINGUO TECH CO LTD
- Filing Date
- 2023-02-15
- Publication Date
- 2026-07-17
AI Technical Summary
Existing image restoration techniques are ill-suited for dental image restoration tasks, mainly due to the high cost of collecting dental image samples and the inability to design effective degradation functions, resulting in poor performance of existing network models with limited paired data.
An improved Auto-Encoder structure is adopted, which combines facial landmark detection and a standard prior encoder. It is trained by L1 Loss, GAN Loss and perceptual Loss functions, and performs dental image restoration using a small number of high-quality dental image datasets. An identity consistency encoder and spatial attention and channel attention modules are introduced to improve the restoration effect.
It enables the generation of realistic tooth restoration effects with very little training data, reduces the difficulty of sample collection and annotation, and improves the adaptability and restoration performance of tooth image restoration.
Smart Images
Figure CN116362995B_ABST