Stereogram Arcade: Applied illusions of depth perception in a video game

Abstract

Stereopsis is the brain’s ability to fuse together the flat and slightly-differing images from each eye into a rich, three-dimensional understanding of space. Autostereograms – images tailored to evoke stereopsis and reveal a 3D scene when viewed with deconverged or “unfocused” eyes – have unique and engaging entertainment value, as demonstrated by the success of the Magic Eye picture book series, but thus far they have been overwhelmingly constrained to the printed, static realm. Autostereographic video games present a unique medium for interactive, animated, 3D entertainment. This thesis has two parts. First, an analysis of game design constraints is presented, with considerations for games intended to be viewed autostereographically, focusing on color and peripheral vision. Second, the design and implementation of a series of autostereogram games that use a custom, real-time autostereogram shader with performance and visual optimizations. This thesis builds upon prior work in developing autostereogram generation algorithms. Prior research discovered several visual optimizations, such as removing false artifacts and directional bias, but still suffered from low-resolution slope discrepancies. Different prior research has achieved real-time rendering speeds using hardware acceleration, but relied on methods that induced visual errors. This game expands on and combines both the performance and visual optimizations. Specifically, slope smoothing and real-time compute shader parallelization are novel optimization techniques for autostereogram generation.