Structural-information-awareness-based regularization model for infrared image stripe noise removal

Infrared images play a crucial role in military reconnaissance, security monitoring, fire detection, and other tasks. However, due to the physical limitations of detectors, an infrared image often suffers from significant stripe noise. The presence of stripe noise significantly degrades image quality and subsequent processing, making the removal of such noise indispensable. In this study, we propose, to our knowledge, a novel low-rank decomposition model to separate the stripe noise components in infrared images. In comparison with existing algorithms for removing infrared stripe noise, our method takes into account the distinctiveness between stripe noise and information components. For the stripe noise component, we describe a column gradient domain low-rank prior and standard deviation weighted group sparsity prior. For the image information component, we employ a structure-aware gradient sparsity prior to suppress stripes while preserving the structural features of images. During the iterative solution process, we utilize both an initial solution based on minimizing column differences and an iteration step-size strategy based on variable acceleration to accelerate convergence. To validate the effectiveness of our proposed method, we conduct experiments to compare it with other destriping algorithms, demonstrating the superiority of our method from the perspectives of both subjective evaluation and objective metrics.