The PL-D755MU-POL is a 5MP camera, with an active area of 2448 x 2048 pixels; but the maximum Region of Interest shown in Pixelink Capture, or when using the SDK, is 1224 x 1024. This is because the sensor pixels are grouped into calculation units, or "polar pixels", as explained below; Creating the effective ROI of 1224 x 1024.


The IMX250MZR, the sensor used by the PL-D755MU-POL, has four directional wire grid polarizing filters formed directly on the sensor’s pixel array.


                         From: https://www.sony-semicon.co.jp/products/common/pdf/IMX250_253MZR_MYR_Flyer_en.pdf


Each sensor pixel is covered by a  0°, 45°, 90°, or 135° polarizing filter. These filters reflect and absorb light waves travelling at an angle parallel to the lines of the wire grid (aka the plane of the polarizer), and allow light waves travelling at an angle perpendicular to the lines of the wire grid to pass through; The pixel then absorbs the linearly polarized light that has passed through it's polarizing filter.


As each sensor pixel is able to detect light from a single polar channel, four sensor pixels, one for each polar channel, are converted into one polar pixel, as shown in the image below. This allows the camera to calculate polarization information like the degree of linear polarization (DoLP) and angle of polarization, and creates the effective ROI of 1224 x 1024.



For more information on the polar pixel formats. see: Image Data Formats


Note: The information from all four sensor pixels is used in each polar pixel; When comparing between the polar and non polar versions of the PL-D755MU, each polar pixel is 48-bits, compared to the 12-bit pixels of the PL-D755MU (when using a 12-bit pixel format). As a result, the same amount of image data is output from both cameras, despite the difference in ROI between the cameras. 


If you are developing your own application and would like to access the individual pixel values for each of the four polar channels, the dotNet PolarFourQuadrant sample application, included with the SDK, provides an example of how to "unpack" the image array to do so.