In this paper, the authors improve the resolution of shortwave infrared (SWIR) polarimetric images using visible images. SWIR sensors typically have larger pixels than their visible counterparts, which limits the resolution at longer wavelengths. The proposed method involves upsampling the SWIR image by six times and essentially reweighing the new pixels based on the high-resolution visible image. An important consideration is that the transition between pixels should be smooth and continuous. To account for this, gradients are calculated between pixels. The paper describes how to take the ratio of these gradients to apply a nonlinear transformation to the reconstructed image, so that the result looks smoother and more natural. As an experimental demonstration, the degree and angle of polarization are computed from high-resolution SWIR images of a coin and teapot. The proposed method can be used for fusing images with different spatial resolutions, as demonstrated for SWIR polarimetry.
You must log in
to add comments.