Rapid advances in semiconductor industry complementary metal oxide semiconductor (CMOS) fabrication process have enabled the integration of a camera onto a single silicon chip. Referred to as CMOS image sensor, a camera-on-a-chip offers significant cant advantages in terms of system miniaturization and manufacturing cost. Like the vast majority of commercially available camera systems, CMOS image sensors are essentially designed to image the world in terms of intensity and color. However, they are "blind" to the third characteristic of light that is polarization. Capturing the polarization state of light reflected or emitted by objects in a scene can not only reveal valuable information about their geometrical, physical, chemical, physiological, and/or metabolic properties but also...[ Read more ]

Rapid advances in semiconductor industry complementary metal oxide semiconductor (CMOS) fabrication process have enabled the integration of a camera onto a single silicon chip. Referred to as CMOS image sensor, a camera-on-a-chip offers significant cant advantages in terms of system miniaturization and manufacturing cost. Like the vast majority of commercially available camera systems, CMOS image sensors are essentially designed to image the world in terms of intensity and color. However, they are "blind" to the third characteristic of light that is polarization. Capturing the polarization state of light reflected or emitted by objects in a scene can not only reveal valuable information about their geometrical, physical, chemical, physiological, and/or metabolic properties but also greatly simplify a number of important machine vision tasks such as autonomous navigation or automatic target detection.

Commercially available CMOS cameras can be made polarization-sensitive by coating individual pixels with micrometer-scale polarizer elements. A micropolarizer array (MPA) is thus formed on top of the photosensitive pixel array. Previously reported micropolarizer implementations suffer from two main limitations: (i) use of selective etching to pattern polarizer elements at the pixel pitch, which is a difficult process to control at the micrometer scale, and (ii) incomplete polarization signature since the circular polarization component cannot be extracted. This circularly polarized component is crucial to fully determine the polarization state of light since, in general, a ray will be elliptically polarized, that is, it can be decomposed into a linear and circular polarization vector. This thesis addresses the above limitations to enable the development of miniature CMOS single-chip polarization cameras capable of capturing simultaneously all possible polarization states of light reflected or emitted by objects in a scene.

This thesis investigates and demonstrates a number of novel MPA design and fabrication technologies, which completely remove the need for selective etching and enable the extraction of all possible polarization states of incident light. The well-controlled-process of ultraviolet (UV) photolithography is used, for the first time, for the patterning of high resolution pixel-level MPAs with submicron thicknesses. Another unique feature of the developed MPAs is that they can be fabricated using standard and well-established CMOS manufacturing flows, enabling low-cost mass production. Both absorptive-type and reflective-type linear polarizers are explored for applications with both natural passive light source (e.g. polarization difference imaging and partial-linear polarization imaging) and artificial collimated active light source for full Stokes imaging. The proposed implementations significantly scale the MPA pitch size down to 2μm. Reported experimental results validate the concept of high performance photo-aligned MPAs with major principal transmittance and extinction ratio as high as ~80% and ~3200 (35dB), respectively.