Introduction to Polarized Cameras

Stefan Waizmann, Technical Marketing Engineer, SVS-Vistek

Along with intensity, propagation direction, frequency, or wavelength spectrum, polarization is one of the basic properties of light. In machine vision, polarization cameras are used to capture polarization image data, which cannot be detected by conventional imaging devices. In doing so, these cameras have opened the door to novel applications in industrial and scientific applications and beyond. Prominent examples include analyzing carbon fiber orientation, visualizing tensions in glass caused by stress-induced birefringence, reducing reflections and glare, or simply enhancing the contrast between materials that are difficult to tell apart with conventional imaging modalities.

What is Light Polarization?

Machine vision applications often rely on automatically analyzing digital images. The images themselves contain information on the interaction of light with the materials of interest. The type of information captured in the images depends on the image acquisition technology employed. A digital monochrome or color camera in combination with adequate optics is most commonly used in industrial machine vision. The cameras capture the intensity distribution and, in case of the color camera, additional information on the wavelength of the reflected or transmitted light of the objects of interest.

Light carries additional information that can be used to probe target objects. The light’s electromagnetic waves are characterized by their intensity, wavelength, phase, and polarization. With polarization cameras, information on the spatial distribution of the polarization state of light can be captured.

Polarization is a property that characterizes the geometrical orientation of the oscillation of the electric field of light. Linear polarized light, for example, is light that oscillates only in one plane perpendicular to the direction of propagation. Graphic A visualizes a linearly polarized wave of one wavelength. Additionally, the figure shows circular polarization.

Linear polarized light can be thought of as the addition of two coherent waves along two orthogonal directions. An optical active media can show a different refractive index for polarization along these two directions such that the two partial waves propagate with different speeds through the media and acquire a phase difference. Adding these two phase-shifted waves results in elliptical or circular polarization, as the resulting electric field vector turns in an elliptic or circular pattern (Graphic B). Circular polarization is obtained for a phase shift of exactly ±90°.

Figures 1A and 1B: (A) Left: Linear polarization at 125°. Right: 2D representation of the electric field vector for a complete cycle. (B) Left: Circular polarization. Right: The 2D representation of the electric field vector for a complete cycle describes a circle. (Images courtesy of SVS-Vistek unless otherwise noted.)

Figures 1C and 1D: (C) Left: Light from multiple sources, as obtained from the sun or incandescent light bulbs, consists of multiple wave trains. They generally show intensity and oscillation orientations that are statistically distributed, so called unpolarized light. Right: Reflection at the interface of dielectric material. The reflectivity of the polarization component that is parallel to the surface normal differs from the polarization component that is perpendicular to it. Thus, the resulting polarization is not unpolarized anymore but partially polarized. (D) For a specific angle that is material dependent, only the polarization component that is parallel to the surface normal gets reflected. This angle is called the Brewster angle.

Measuring the Polarization

Linear polarizers can be used to characterize the light polarization. It is an easy and efficient way to obtain linear polarization from unpolarized light. Multiple realizations of polarizers exist and explaining them all would surpass the scope of this article. Here we want to focus on the grid polarizer. This is an optical element consisting of an array of subwavelength parallel metal nanowires. The component of the incident electric field that is polarized parallel to these metal wires is blocked in a way that only the polarization component perpendicular to the nanowire grid is transmitted. These nanowire grids are employed in some polarized cameras.

Figure 6: The SVS-Vistek exo250Z camera uses the 5 MPixel IMX250MZR Sony sensor. The Sony sensor is based on the popular 2448 x 2049 pixel, 2/3” IMX250 CMOS sensor with 3.45 μm pixel size.

For example, think about, a camera using the 5 MPixel IMX250MZR Sony sensor. The Sony sensor is based on the 2448 × 2049 pixel, 2/3” IMX250 CMOS sensor with 3.45 μm pixel size. A four-directional polarization square filter array is overlaid directly on top of the pixel array and beneath the micro lenses. This filter consists of repeated 2×2 patterns consisting of grid polarizers with four different angles at 0°, 45°, 135°, and 90°. Each polarizer filters the incoming light so that only the polarization components perpendicular to the grid orientation can pass through and be detected by the underlying photodiode. Thus, a four directional polarization image can be captured in one shot.

Figure 2: A grid polarizer blocks the polarization component that is parallel to the grid array. Only light perpendicular to the grid can pass through.

Figure 3: Grid polarizers with different grid orientations result in linear polarized light with orientations perpendicular to the grid orientation, respectively. Here, grid orientations of 0°, 45°, 135°, and 90° are shown.

Figures 4A and 4B: (A) The Sony IMX250MZR sensor has a polarization filter array added to the photodiode array. A short distance between polarizer and the photodiode reduces the effect of crosstalk, i.e., the wrong detection of a polarization angle by a neighboring pixel. The sensor shows excellent image quality in various light source environments. (B) The polarization filter array consists of multiple 2x2 patterns that show four different wire grid orientations. The respective measured signal I of a 2x2 pattern is a measure of the amount of light with 0° (I0), 135° (I135), 45° (I45), and 90° (I90) polarization. These intensities can be used to estimate the angle of linear polarization as well as the DoLP. (Images are adapted from Sony)

In subsequent post processing, the 5 MPixel image can be used to obtain four size-reduced images (1.25 MPixels) filtered by one polarization filter, respectively. Additionally, the images can be used to estimate the spatial distribution of the brightness of the average of each 2x2 pattern, as well as the associated linear polarization angle and the DoLP.

The pixel intensities of a 2×2 pattern are a measure of the amount of light with 90° (I₉₀), 135° (I₁₃₅), 45° (I₄₅), and 0° (I₀) polarization, respectively. Addition and subtraction of these intensities are known as Stokes Parameters. Note that the fourth Stokes parameter S3 = IR – IL, which is calculated from the intensities of right and left circular polarized light, cannot be obtained from the image.

S₀=I₀ + I₉₀

S₁=I₀ – I₉₀

S₂=I₄₅ – I₁₃₅

Conveniently, the polarization angle and the DoLPare easily computed from the Stokes parameters. The inverse tangent needs to be computed such that it returns values in the closed interval [-180°,180°].

Examining these values provides imaging contrast enhancements in a multitude of applications compared with standard camera imaging. Also, note that the degree of linear polarization does not quantify the amount of circular polarized light. Perfectly circular polarized light will have a value of DoLP = 0. Nevertheless, many applications do not require estimating the circular polarization.

Figure 5:Angle of polarization as compared to the orientation of the 2x2 four-directional polarization pattern. DoLP is zero for unpolarized light and one for perfectly linear polarized light. In the case of light with partial polarization, a value between zero and one is obtained.

Polarization Application Examples

Calculating the spatial distribution of the polarization angle as well as the DoLP increases the information content of the measurement.

Many applications exist for such measurements—for example, analyzing the orientation of carbon fibers, visualizing tensions in glass caused by stress induced birefringence, reducing reflections and glare, or enhancing contrast between similar materials that are difficult to tell apart with conventional imaging methods.

Some application examples are shown.

Carbon Fiber Example: Monochrome camera image (left). The polarization angle allows the estimation of the carbon fiber orientation (right). Additionally, potential errors in a carbon fiber grid can easily be detected.

Monochrome image of a pack of pills (left). The aluminum pack reflects the linear polarized light, whereas the reflection from the pills results in unpolarized light. Thus, the DoLP shows a strong contrast between the pill and the package. The missing pill can easily be identified (right).

Cell phone LCD, 10x magnification and illuminated with polarized red light (left). The LCD pixels with red color filter appear brighter. The red filter reflects the light directly; The blue and green filters let part of the light pass through. Interaction with the underlying optical active media induces a measurable change in the polarization state (right).

Plastics, glass, metals, and liquids display intrinsic polarization properties that result in unwanted glare and other artifacts using traditional imaging. Polarization cameras use polarization properties to visualize defects such as stress and scratches, while reducing glare, improving edge detection, and enhancing contrast. By filtering light, polarization cameras will improve a machine vision system’s capture, especially when inspecting objects that would otherwise be very difficult to identify in applications as diverse as semiconductors, consumer electronics, flat panel displays, packaging, and microscopy.