Multi-drop attention residual infrared image denoising network based on guided filtering

At present, infrared images are widely used in various fields, but limited by the non-uniformity of detector unit, the infrared image has the disadvantages of low signal-to-noise ratio and blurred visual effects, which seriously affect its application in advanced fields. Commonly used denoising algorithms cannot take into account the smoothing of denoising and the preservation of edge details. In response to the above problems, this paper proposes a multi-drop attention residual denoising network based on guided filtering. A guided convolution module is designed according to the principle of guided filtering and a multi-drop attention residual module is designed for both the extraction of shallow and deep features. Experiments have proved that the network after adding the new module can effectively reduce the noise of infrared images, and can maintain the edge detail information in the image to the greatest extent, improve the visual effect, and also have good performance on the PSRN and SSIM indicators.