Computational imaging with light waves

Abstract

Modern imaging systems have been evolved to effectively capture light. Through continuous developments, we can now capture the world realistically even competitive to the human visual system. However, capturing light in full fidelity is still challenging. One fundamental reason originates from the fact that light has the properties of both particles and waves. Most imaging devices are not optimized to capture light as waves, therefore, losing valuable information about the real world hidden in the wave properties of light. This dissertation studies ways of capturing, analyzing and exploiting the overlooked dimensions of light waves, such as spectrum and polarization, in order to solve many problems in computer vision and graphics. To this end, we develop imaging systems so that the wave properties of light can be captured in the form of images. Computational algorithms then extract scene properties from the captured images. This joint design of optics and computational algorithms enables us to understand the complex and delicate interaction of light waves and materials. Specifically, we demonstrated four applications of this principle: estimating depth from double refraction, color and depth from uneven double refraction, spectrum from dispersion, and surface appearance from polarization. We believe that computational imaging systems of capturing light waves in full fidelity will open up the completely new understanding of our real world.