top of page

Image reconstruction using polarized light

In one of my research rotations, I developed a proof of concept showing that information currently not registered by optical sensors (cameras) has the potential to render better images. Immediate applications that come to mind are cameras for surgery and security.

Polarization sensors Research

Regular cameras capture only natural light to form an image. Nevertheless, more information is present in the polarized part of the light. The study suggests that polarization sensors form qualitatively better images under lighting conditions where the human eye or regular cameras cannot form any decipherable image. Furthermore, we often get an even better image of the object using superposition of the angle of polarization with the degree of polarization.

​​​

Consider the two strips of pictures below. The first (leftmost) photo in each strip was taken with a regular camera (Canon Rebel T3). The rest of the images were taken with the sensor designed and built in the lab, which captures the data about the polarization of light. The second image from the left superimposes the degree of polarization of light reflected from an object. In the third image, we superimpose the angle of polarization. Finally, in the rightmost image, we reconstruct the image by superimposing both the degree and angle of polarization.

An image taken with a regular camera.

Image reconstructed from the degree of polarization.

Image reconstructed from the angle of polarization.

Image reconstructed by superimposing the degree and angle of polarization.

As can be seen, the rightmost picture above gives a human observer enough details to be perceivable, almost to the point that the picture could be mistaken for taken in daylight.

The purpose of the next experiment is to question flat object vision (objects that do not have 3D shapes), in this case, the ability to use polarized light for text reconstruction. Using the same equipment and image reconstruction algorithm, we took a picture of a cup with a textual logo. From left to right, we show the progression of the image reconstruction in the same manner discussed above.

Image2 -- Canon Rebel T3.png

An image taken with a regular camera.

Image2 -- DOP.png

Image reconstructed from the degree of polarization.

Image2 -- AOP.png

Image reconstructed from the angle of polarization.

Image2 -- DOP and AOP.png

Image reconstructed by superimposing the degree and angle of polarization.

In the rightmost picture, even the small pores of the bricks from which our lab wall (the wall behind the cup) is built are perceivable.

Mentor

  • Dr. Viktor Gruev

Resources

bottom of page