IRFPA strong nonuniformity adaptive correction algorithm based on adjacent pixel differential statistics

Uncooled infrared cameras may exhibit pronounced nonuniform noise, especially during extended operation without recalibration, or under typical gas leak detection narrow-band imaging conditions. Traditional nonuniformity correction methods focus on removing high-frequency noise, which is ineffective for such strong fixed-pattern noise (FPN) with wide-band characteristics. This study investigates a strong nonuniformity adaptive correction algorithm based on adjacent pixel differential statistics and independent of noise bandwidth. Based on the frequency characteristics of FPN, a progressive multistage correction strategy was adopted to effectively separate strong FPN from scene information by exploiting its separability in the temporal and spatial domains. The error propagation model of the adjacent pixel differential statistical method was derived, forming the basis for a spatiotemporal adaptive ghosting suppression approach under complex noise conditions. This process enabled both the adaptive correction of strong FPN and the real-time update of detector drift correction parameters. Simulated and real-image experiments demonstrate that in strong FPN correction, the proposed algorithm offers strong adaptability, superior correction performance, improved subjective visual quality, and better objective evaluation metrics compared to other advanced methods while maintaining low computational cost. The proposed algorithm shows great potential in diverse applications in image nonuniformity correction.