Image region missing content generation method and system based on three-dimensional partial differential equation
By combining three-dimensional partial differential equations and deep learning, the progressive image generation network PDE-PINet solves the problem of poor structure and detail in the generation of missing content in image regions, and achieves more natural and detailed image generation results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING UNIV OF CIVIL ENG & ARCHITECTURE
- Filing Date
- 2025-10-29
- Publication Date
- 2026-07-24
AI Technical Summary
Existing technologies suffer from problems such as easily confused image structure and loss of details in the generation of missing content in image regions, especially when generating complex textures and detailed images. Furthermore, deep learning methods lack rigorous mathematical foundations and utilization of frequency domain information.
We employ a progressive image region missing content generation network, PDE-PINet, based on three-dimensional partial differential equations. Combining partial differential equations with deep learning, we utilize a multi-stage generation strategy and a frequency domain attention mechanism to drive image generation using three-dimensional partial differential equations. By combining semantic segmentation maps and multi-scale feature extraction, we can achieve the generation of global structure and local details of images.
It maintains the overall structural coherence of the image in the generation of large missing regions, and the generated image has natural texture transitions and rich details, which improves the generation quality. It is suitable for complex texture and detailed images and has better generation effect and cross-dataset generalization ability.
Smart Images

Figure CN121544733B_ABST