Adaptive strategy for demosaicing microgrid polarimeter imagery

Abstract

Microgrid imaging polarimeters consist of a focal plane array sensor with linear polarization filters of differing orientations overlaid at each pixel, similar in concept to the arrangement of spectral filters in a color CCD Bayer pattern camera. However, unlike spectral color cameras, microgrid systems use polarimetrically modulated intensity measurements to reconstruct the Stokes vector at each point in an imaged scene. Stokes reconstruction of imagery from these devices has traditionally been performed using linear filtering techniques. While linear filtering strategies can yield reasonable estimates of the Stokes imagery, the filtering often results in loss of high frequency content in addition to introducing typical demosaicing artifacts (such as aliasing and zippering effects). Here we develop an adaptive demosaicing strategy based upon the concept of bilateral filtering as a means for reducing interpolation artifacts while preserving high frequency image content often removed by non-adaptive linear interpolators. We demonstrate the demosaicing strategy and compare it against imagery estimated using other techniques on LWIR microgrid data.