Stereoscopic 3D provides an immersive viewing experience and has evoked a tremendous amount of interest. However, stereoscopic content must be properly produced to avoid "3D fatigue," such as blurring vision, eyestrain, and headache. The difficulty in producing stereoscopic content has become a bottleneck that limits its impact.
This research develops computational stereoscopic cinematography technologies to assist stereoscopic content production by integrating computer graphics, computer vision, stereoscopic cinematography, and stereo perception. This project explores fundamental and unique problems in computational stereoscopic cinematography: how to edit disparity to deliver a pleasant viewing experience; how disparity interacts with other aspects of stereoscopic content like monocular depth cues and motion; how to optimize disparity during stereoscopic content manipulation and authoring, including warping, collage authoring, and video stabilization; how to coordinate the process of the left and right view with no or unreliable stereo correspondences; and how to extend monocular content manipulation and authoring technologies to stereoscopic content in a principled way. In short, this project solves key problems in computational stereoscopic cinematography and establishes general principles for developing stereoscopic content processing and authoring technologies.
This computational stereoscopic cinematography research facilitates high-quality stereoscopic content production, which is critical in sustaining and boosting the impact of stereoscopic 3D in a variety of applications, including virtual reality, human computer interaction, scientific visualization, educational training, and medical rehabilitation. The research results, including the technical writings, the stereoscopic image/video library, and the code infrastructure, are disseminated through paper publications and online sharing.
