一种补偿互联网传输降质的人脸图像增强方法
By designing a face image enhancement model that incorporates shallow feature extraction, deep feature extraction, and subpixel convolutional reconstruction, the problem of face image degradation during internet transmission is solved, achieving efficient and realistic image reconstruction results suitable for mobile devices.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TIANJIN UNIV
- Filing Date
- 2024-10-31
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, facial images are easily degraded during internet transmission due to low resolution, blurring, and compression artifacts, resulting in poor image reconstruction quality and low computational efficiency, making them difficult to deploy on mobile platforms.
A face image enhancement model is designed, consisting of shallow feature extraction, deep feature extraction, and subpixel convolutional reconstruction. It utilizes convolutional neural networks and residual convolutional neural networks for feature extraction and subpixel convolutional networks for image reconstruction. By combining the skip connection method of window multi-head self-attention mechanism and sliding window multi-head self-attention mechanism, the computational complexity is reduced.
It achieves high-quality face image reconstruction with realistic results, few artifacts, and low computational cost. It is suitable for mobile devices, effectively improves image resolution while preserving detail information, and is applicable to face image enhancement in Internet applications.
Smart Images
Figure CN119991473B_ABST