Multi-band polarization imaging, by mean of analyzing spectral and polarimetric data simultaneously, is a good way to improve the quantity and quality of information recovered from a scene. Therefore, it can enhance computer vision algorithms as it permits to recover more statistical information about a surface than color imaging. This work presents a database of polarimetric and multispectral images that combine visible and near-infrared (NIR) information. An experimental setup is built around a dual-sensor camera. Multispectral images are reconstructed from the dual-RGB method. The polarimetric feature is achieved using rotating linear polarization filters in front of the camera at four different angles (0, 45, 90 and 135 degrees). The resulting imaging system outputs 6 spectral/polarimetric channels. We demonstrate 10 different scenes composed of several materials like color checker, high reflecting metallic object, plastic, painting, liquid, fabric and food. Our database of images is provided online as supplementary material for further simulation and data analysis. This work also discusses several issues about the multi-band imaging technique described.
This work shows the interest of combining polarimetric and light-field imaging. Polarimetric imaging is known for its capabilities to highlight and reveal contrasts or surfaces that are not visible in standard intensity images. This imaging mode requires to capture multiple images with a set of different polarimetric filters. The images can either be captured by a temporal or spatial multiplexing, depending on the polarimeter model used. On the other hand, light-field imaging, which is categorized in the field of computational imaging, is also based on a combination of images that allows to extract 3D information about the scene. In this case, images are either acquired with a camera array, or with a multi-view camera such as a plenoptic camera. One of the major interests of a light-field camera is its capability to produce different kind of images, such as sub-aperture images used to compute depth images, full focus images or images refocused at a specific distance used to detect defects for instance. In this paper, we show that refocused images of a light-field camera can also be computed in the context of polarimetric imaging. The 3D information contained in the refocused images can be combined with the linear degree of polarization and can be obtained with an unique device in one acquisition. An example illustrates how these two coupled imaging modes are promising, especially for the industrial control and inspection by vision.
For any kind of imaging polarimeter, at least four intensity images, named polarization state images, are needed
to compute one full Stokes vector. When the polarimeter is designed according to the division-of-time principle,
polarization state images are acquired sequentially. Consequently, the main issue is the systematic occurrence of
artifacts as the scene is not static. Even though this is well known, little research has been done on this subject.
Here a two-step motion-compensation-based method is proposed to fix it. The first step consists in estimating
the motion between each image acquired according to the same polarization state. Then each image is warped
according to a fraction of the previously estimated motion.
Due to their dense and accurately estimated motion field we have shown optical-flow techniques to be the most
efficient for motion-estimation in this case. Compensating the motion using optical flow to estimate it actually
leads to a strong correlation criterion between corrected and reference polarization images.
Our method allows the estimation of the polarization by post-processing the polarization state image sequence.
It leads to a good estimation quality whether the scene is static or not, thus fixing the main issue of a divisionof-
This paper reports the design and the implementation of a Stokes imaging polarimeter able to provide full polarimetric information at 200 fps. This portable implementation is based on a division-of-time architecture and uses a single ferroelectric liquid crystal device as the polarization modulating element. Our system is designed to work at 532 nm with natural light or with controlled illumination, without temperature control. We propose an optimized driving scheme of the modulator such that the liquid crystal device can produce four polarization states which makes it possible to retrieve the full polarimetric information. The modulator characterization is reported and experimental results are provided.
We already implemented an imaging polarimeter able to capture at high-speed the full information about linear
polarization of a monochromatic light beam (namely its first three Stokes parameters). The polarizing element
was a single ferroelectric liquid crystal cell, acting as a half-wave plate (therefore as a polarization rotator)
at its design wavelength. In this paper, we report the improvement of our system in order to grab the full
Stokes information, including the fourth Stokes parameter. The procedure consists in two operations. First,
optimally controlling our polarization component with an additional fourth voltage level. Second, shifting the
wavelength operation in order to get benefit of the device chromatic behavior: away from its design wavelength,
the device does not behave as a half-wave plate any longer, and with proper level control, the system matrix of
our polarization state analyzer (consisting of the liquid crystal cell and of a fixed linear polarizer) can reach rank
4. Therefore, elliptical polarization can be fully analyzed, as it could be with a nematic liquid crystal device,
but at a much higher frame rate. Results of operation at 200 fps are provided.
We present a high-speed ferroelectric liquid crystal based imaging polarimeter. It can evaluate the first three
Stokes parameters. Contrary to previous high-speed systems, it only uses a single liquid crystal cell, driven in
an optimized way in order to produce a tunable rotation of polarization. Its characterization is presented, as
well as its integration in a portable implementation working at 633 nm. Preliminary results are provided.
The polarimeter we developed is composed of a single commercial CCD camera with a bistable ferroelectric liquid
crystal modulator as a polarizing element; therefore the degree of polarization (DOP) is evaluated from two successive
acquisitions. Thus, when an object moves during the acquisition, issues occur in the polarization information
reconstructed from the two state pictures: edges of objects have neither the same DOP as the object itself nor the same
DOP as the background. In this paper, we present two methods to correct defects in DOP images of moving objects.
First, we present a post-processing temporal median filtering, correcting the DOP once computed. The second method
consists in performing a motion estimation to correct the object's displacement between the two polarization state
pictures. Comparison based on quantitative results with real data is provided.