一种融合生成对抗网络与不确定性的图像质量增强方法
By introducing uncertainty estimation into generative adversarial networks, the credibility problem of image data caused by black box models is solved, and the credibility and interpretability of image quality enhancement are improved.
CN117745591BActive Publication Date: 2026-07-17UNIV OF ELECTRONICS SCI & TECH OF CHINA
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- UNIV OF ELECTRONICS SCI & TECH OF CHINA
- Filing Date
- 2023-12-22
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing generative adversarial networks suffer from a lack of interpretability in image quality enhancement, resulting in low image data quality and affecting the credibility of applications such as medical image diagnosis.
Method used
Uncertainty estimation is introduced into generative adversarial networks. By constructing a generator and a discriminator, uncertainty is estimated using information theory. An objective function is designed and trained alternately until convergence, thereby improving the credibility of image quality enhancement.
Benefits of technology
By eliminating uncertainties, the training interpretability of generative adversarial networks is improved, and the credibility of image enhancement and the accuracy of prediction results are enhanced.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN117745591B_ABST
Abstract
本发明属于图像视觉和图像增强技术领域，具体提供一种融合生成对抗网络与不确定性的图像质量增强方法，用以提高图像数据质量，进而提高图像数据对预测结果的可信度。本发明在生成对抗网络(GAN网络)中引入不确定性估计，将GAN网络与信息论、主观逻辑等理论相结合，通过在判别器中使用深度证据网络输出图像所属类别的证据，估计GAN网络输出的空度、不协调度、偶然不确定性和认知不确定性，并在生成器的目标函数中引入消除不确定性项，促使生成器的目标朝着消除预测不确定性的方向演化，最终使得GAN网络的训练具备信息论可解释性的同时，提高图像增强的可信度。
Need to check novelty before this filing date? Find Prior Art