Super-resolution Reconstruction of Thermal Images by UAV Based on Anisotropic Filtering

The super-resolution reconstruction method of UAV thermal infrared images based on anisotropic filtering was studied to improve UAV thermal infrared images resolution and provide efficient and accurate identification of UAV inspection images. The adaptive setting was based on the image edge area and smooth area gradient threshold, under the set by the anisotropic diffusion filter method. We aimed to remove the initial thermal infrared image noise smoothing area, maintain details, enhance the image edges, and obtain a thermal infrared image after denoising. To generate an image against the network input, the model and discriminant model were generated by game learning for super-resolution reconstruction of the thermal infrared image output. This method has high image edge detail protection, the ability to remove background noise, and an improved comprehensive denoising effect. Furthermore, it can obtain high-resolution and visual presentation of super-resolution reconstruction of thermal infrared images, as well as ensure the identification accuracy and efficiency of thermal infrared images in UAV inspection