Infrared image detail enhancement algorithm based on Cauchy filtering and multiscale decomposition

With the rapid development of infrared sensors, infrared imaging has gained widespread applications in military, medical, industrial, and intelligent surveillance fields. However, infrared images often suffer from low quality, and traditional enhancement methods may result in detail loss, poor generalization, or inadequate real-time performance. To address these challenges, this project employs, to our knowledge, a novel approach that integrates image decomposition with detail information fusion. Our independently designed Cauchy filter, in conjunction with fast guided filtering, effectively extracts detail information from the source image. A nonlinear function adaptively blends detail weight coefficients while a detail adaptive gain factor is introduced to control the weight of reconstructed details based on the source image's mean and variance. For the background layer, adaptive <i>γ</i> correction is applied to enhance image contrast. Additionally, we propose a noise reduction framework for the reconstructed image, producing infrared images with richer information and improved scene representation. This approach is suitable for various complex scenarios, enhances user perception, and facilitates higher-level semantic tasks.