Ph.D Thesis

Ph.D StudentTreibitz Avital
SubjectGeometry and Photometry of Imaging through a Medium
DepartmentDepartment of Electrical and Computer Engineering
Supervisor PROF. Yoav Schechner
Full Thesis textFull thesis text - English Version


Images taken through a medium may suffer from poor visibility and loss of contrast. Light passing  through undergoes absorption and scattering. Wavelength dependent attenuation causes changes in color and brightness. In addition, light that is scattered back from the medium into the camera (\emph{backscatter}) veils the object, degrading visibility and contrast. Moreover, refraction between the  medium and the camera (in air) causes geometric distortions that harm geometric reconstruction. Nevertheless, there is a strong need to perform vision tasks in such media. Thus, in this work, we look both at photometric and geometrical aspects of imaging in these conditions.

In photometry, we first show that backscatter can be significantly reduced by post-processing polarized frames in active illumination. Our method greatly enhances visibility and also enables rough estimation of the 3D scene structure. However, it is limited in range, since the signal-to-noise-ratio (SNR) falls sharply with distance. Consequently, we study how noise imposes limits on recovery, when images are degraded by a pointwise process, such as reflections, direct and indirect illumination components and scattering media. We show that in pointwise degradations noise imposes resolution limits. The result is a tool that assesses the ability to recover (within a desirable success rate) an object or feature having a certain size, distance from the camera, and radiance difference from its nearby background, per attenuation coefficient of the medium. The bounds rely on the camera specifications.

In addition, we analyze tradeoffs in polarization-based visibility enhancement methods. Mounting a polarizer might improve visibility but it also attenuates the signal associated with the object. This attenuation degrades the image quality. Thus, a question arises: is the use of a polarizer worth the mentioned loss? Our analysis shows that polarization rarely improves the SNR, for a fixed acquisition time.

In geometry, we study a common camera configuration, which looks into water through a flat window, e.g., an underwater camera; imaging into an aquarium; and looking into water through goggles or a diving mask. We show that such a configuration does not have a single viewpoint, due to refraction. However, previous works in this domain used calibration schemes for single-viewpoint systems. Therefore, methods that aim to find a perspective transformation should fail, as indeed they do. We thus develop a calibration method tailored to this common configuration.