Infrared non-uniformity correction algorithm based on classification and regression tree segmentation

In infrared imaging, pixel non-uniformity due to manufacturing and technological limitations significantly degrades image quality, posing a critical challenge for high-performance applications. This issue is especially pronounced in the field of infrared astronomical observation, where the scientific integrity of the data must be ensured when correcting infrared images. In addition, the coupling capacitance between pixels leads to nonlinear effects in pixel sensitivity. To address these challenges, a non-uniformity correction (NUC) algorithm was introduced that utilizes classification and regression tree segmentation. This approach enables precise corrections by adapting to varying illuminance levels, a capability not fully explored in existing solutions. In our method, we innovatively segment the pixel response curves into distinct low and high illuminance ranges and apply customized corrections for each segment to enhance the accuracy of correction. Evaluations using real image data demonstrate that our method enhances image quality and consistency.