Skip to Main Content
Spatial characteristics of thermal infrared (TIR) images are generally described by image-derived point spread function (IDPSF) for sampled imaging system. Most investigators working on IDPSF on-orbit evaluation focus on visible and near-infrared images and pay little attention on TIR ones, the reason of which lies on the fact that the assumption of an observed target with an ideal step profile is not suitable for TIR band. A new approach for TIR images is proposed here based on a more common slope-profile model, where the transition features, i.e., the sample number (N slope) and the beginning position of slope interval, are extracted from the observed signals. Simulation results show that the estimation performance will be increased, with lower noise level as well as higher system modulation transfer function value at half of normalized spatial frequency (MTF 0.5) or smaller N slope. For a good performance system with noise squared-root variance (SRV) no more than 1.0 count and MTF 0.5 between 0.3 and 0.4, the relative error between simulated and estimated MTF 0.5's is less than 20%. The IDPSF of FY-2C satellite TIR band is estimated, and the derived MTF 0.5 is properly close to on-ground testing results. The image quality of FY-2C TIR band has been improved by using Wiener filtering and the estimated IDPSF, which eventually benefits the cloud detection product with more detectable low fractus and increased detection accuracy by about 1% in winter. Considering a step-profile target as a special case of this model, the proposed approach is also suitable for visible and near-infrared images.