Radiometric imaging by double exposure and gain calibration.

Abstract

Due to low cost and small size, uncooled microbolometer-based thermal focal plane arrays are very attractive for radiometry. However, being non-cooled, they suffer from temporally and spatially dependent changes that require constant calibration. While the gain calibration can be reasonably realized by two-point correction, the offset due to internal radiation loads poses a complicated calibration scheme. We present a new computational optics approach that simplifies the essential calibration for temperature offset. Using two successive images of the object taken with different known blur levels, one can eliminate the object term from the image-formation equation, resulting in an equation for the unknown sensor offset. A general algebraic model is presented for the space-variant case followed by solutions using both direct inverse method and iterative solver. The new scheme allows restoration of the radiometric value within 1% error with the direct method, and 0.2% error with the iterative scheme. Account of the influence of realistic lens positioning error on restoration accuracy was given. Results using direct inverse methods for restoring the radiometric values yield restoration error with a good average error of 3.7% and less.