Assessing and deconvolving the impacts of the point spread function on satellite remote sensing

Abstract

The point spread function (PSF) describes the response of sensor systems to point signals. The quality of satellite images is affected by this function. In this study an experiment was designed to quantitatively assess the impacts of the scan mirror's motion's PSF on images and subsequently derived land cover products from the Moderate Resolution Imaging Spectroradiometer (MODIS). The scan mirror's motion causes about 25% of the radiance of a pixel to contribute to its neighboring pixels in the scan direction, half to the left and half to the right. On average this factor alone can result in deviations of up to 1.2 digital number (DN) of satellite signals from those not affected by the PSF, corresponding to 1.8-3.5% of the DN range of the corresponding bands. The maximum deviation can be up to 15% of the DN ranges for the visible and near infrared bands. The percentages of subpixel land cover were derived from the PSF impacted image using a linear mixture model. Errors of similar magnitudes were found when the predicted percentages were compared to the actual values. A deconvolution method was developed based on the PSF model. The impacts of the PSF on both images and derived land cover percentages were reduced by about 30% using the developed deconvolution method.