A correlation-based interpolation algorithm for division-of-focal-plane polarization sensors

We propose an interpolation algorithm for Division-of-Focal-Plane (DoFP) polarimeters based on the correlation between neighboring pixels. DoFP polarimeters monolithically integrate pixelated nanowire polarization filters with an array of imaging elements. DoFP sensors have been realized in the visible and near-infrared regime. The advantages of DoFP sensors are twofold. First, they capture polarization information at every frame. Second, they are compact and robust. The main disadvantage is the loss of spatial resolution due to the super-pixel sampling paradigm at the focal plane. These sensors produce four low-resolution images, where each image has been recorded by a linear polarization filter offset by 45 degrees. Our algorithm addresses the loss of spatial resolution by utilizing the correlation information between the four polarization pixels in a super-pixel configuration. The method is based on the following premise: if one or more of three polarization parameters (angle of polarization, degree of polarization, and intensity) are known for a spatial neighborhood, then the unknown pixel values for the 0° image, for example, can be computed from the intensity values from the 45°, 90° and 135° images. The proposed algorithm is applied to select cases and found to outperform the bicubic spline interpolation method.