190 / 2019-12-27 11:28:00

Cross-track Illumination Correction For Hyperspectral Pushbroom Sensors Using Total Variation and Sparsity Regularization

Hyperspectral imaging

IEEE SAM 2020: Special Sessions > Unsupervised Computing and Large-Scale Optimization for Multi-Dimensional Data Processing

Cross-track illumination error exists in hyperspectral pushbroom sensor, who scan objects line-by-line with a detector array. When the illumination sensitivity of the individual detectors is not aligned well, or some detectors are degraded/aged, acquired images show non-uniform illumination in the cross-track direction. Meanwhile, because of the line-by-line scanning scheme, the cross-track illumination error is replicated along the flying track. Considering the structure of illumination error cross/along the track, we propose a column (along-track) mean compensation approach with total variation and sparsity regularization (COMCO-TVS), which corrects the illumination via exploiting characteristics of column-mean pixels and column-mean illumination errors: piecewise smoothness and sparsity, respectively, in the spatial-spectral domain. The correction effectiveness of the proposed method is illustrated using semi-real data.