单帧红外热图像增强方法及系统
By simultaneously acquiring infrared thermal images and visible light images, calculating virtual diffusion coefficients, generating Gaussian diffusion kernels and convolution kernels, performing convolution and frequency domain transformations, and constructing a conditional generative adversarial network, the problem of noise interference and image defocusing in single-frame infrared images under complex scenes is solved, achieving high-precision thermal flux assessment and enhancement effects.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUNAN UNIV
- Filing Date
- 2026-04-29
- Publication Date
- 2026-07-17
AI Technical Summary
Existing single-frame infrared image enhancement techniques are difficult to achieve high-precision heat flux assessment in complex scenarios. They suffer from severe noise interference, limited image quality, difficulty in balancing physical models and data-driven methods, and severe image defocusing in multispectral detection, failing to meet the requirements for high-precision detection.
By simultaneously acquiring infrared thermal images and visible light images, calculating virtual diffusion coefficients, generating Gaussian diffusion kernels and convolution kernels, performing convolution operations and frequency domain transformations, constructing a conditional generative adversarial network, incorporating multispectral scene conditional information, training the network with a dual loss function, and outputting a target-enhanced thermal flow image.
It improves the signal-to-noise ratio of the image, eliminates halo artifacts, enhances the physical rationality of heat flow inversion, improves the clarity of defect edges and the reliability of detection results, and meets the requirements of high-precision quantitative heat flow analysis.
Smart Images
Figure CN122134588B_ABST