İmge Kıymetlendirmede Polarizasyon Tekniklerinin Kullanımı
Özet
Polarization provides more detailed description of the light than its energy. For this
reason, polarized images contain more information than conventional images.
Polarization is not noticeable by the naked eye by most living things, but can be
detected by additional auxiliary materials such as polarized filters.
The investigations of polarized light began with the discovery of the phenomenon
of double refraction in calcite crystals by Erasmus Bartholinus (1625–1698) in
1669. With foundation of the wave equation by James Clerk Maxwell’s (1831–
1879) electrodynamic theory and its experimental confirmation by Heinrich Hertz
(1857–1894) in the second half of the nineteenth century, polarization has its
mathematical bases proven with experiments.
Polarization images can be obtained with a simple mechanism by putting a
polarizer in front of the cameras. The use of polarization in computer vision has
begun to become widespread in recent years, as the simplicity of obtaining
polarized images through cameras and the fact that polarized images contain
more information than traditional images. Polarization is used many areas like
iv
detecting flat surfaces, finding surface orientation angles, detecting conductive or
insulating objects, increasing the view distance due to the haze effects, creating
depth maps, removing semi-reflections in the transparent surfaces, edge detection
and constructing 3D shapes of objects.
In this study, after definition of polarization, it has been shown that
parameterization can be done with Stokes parameters by using three different
measurements taken at 0°, 45° and 90° angles of polarized filter. After that, the
use of polarization in computer vision has been investigated and related studies
have been collected under four sections.
A test set-up was obtained to take polarized images and various field studies were
carried out. An image registration technique has been applied to correct the
distortions in the polarized images obtained in a sequential manner with the
related test set-up.
In the experimental studies, firstly polarized imaging is emphasized and studies
such as segmentation of various surfaces with polarization and increasing of
visibility are presented. After that, it has been shown that semi-reflections
occurring on transparent or semi-transparent surfaces can be separated by
independent components analysis. The performance of the relevant
decomposition method has been analysed in various scenarios. It also has been
shown that the visibility could be increased by using polarization and independent
component analysis when haze occurred in the scene. In addition to this study, the
sky polarization characteristics has been investigated. As a result of this, it has
been shown that the degree of linear polarization and angle of linear polarization
are depend on celestial position of the sun. The degree of linear polarization of a
light beam coming from clouds is less than the clear sky.