Impact of Multiple Variable Integration Times on Single-FrameNon-Uniformity Correction : A Study Using a Mid-Wave Infrared Camera Performed at Saab Dynamics AB

Sammanfattning

This thesis analyses how multiple integration times can be used for single-frame non-uniformity correction to reduce fixed pattern noise in images captured with a cooled mid-wave infrared camera. Two non-uniformity correction methods are implemented: a calibration-based approach using two integration time presets, and a scene-based approach based on the discrete wavelet transform. The two methods are evaluated and compared using image sequences mainly containing blackbody reference sources and static scenes. To investigate the influence and potential drawbacks of using multiple integration times, the calibration-based approach is further analysed for moving scenes while varying the shorter integration time. In addition, three variants of the baseline methods are implemented and evaluated to address limitations in one or both approaches. The evaluation combines quantitative metrics, measuring non-uniformity and image quality after correction, with visual assessment. The results show that the calibration-based integration time approach effectively reduces both high-frequency stripe noise and low-frequency non-uniformity, but that its performance depends strongly on the shorter integration time and is sensitive to motion between frames. The discrete wavelet transform-based approach also reduces stripe noise to some extent, but may introduce artefacts when image structures are difficult to separate from noise. The modified variants show that some limitations can be compensated for, and that information from multiple integration times can be useful for improving single-frame non-uniformity correction.