Shadow-Free Augmented Reality Makes Illusions More Realistic

Researchers at the University of Osaka have built a multi-projector augmented reality system that suppresses shadows and corrects blur, producing projections that are harder to distinguish from real printed surfaces, according to a study published in May. The system, described in IEEE Transactions on Visualization and Computer Graphics, uses multiple projectors aimed from different angles so that if a user’s hand blocks one beam, light from the remaining units still reaches the surface [1]. Projection augmented models have long relied on overlaying computer images onto physical objects to alter their perceived color or texture [2], but overlapping images from separate projectors typically create blur because each projector illuminates the surface from a slightly different direction [1]. Daisuke Iwai, a professor at the University of Osaka’s Graduate School of Engineering Science, said the team addressed this by estimating the combined blur and generating a single pre-compensated image shared across all projectors after geometric alignment. “This keeps the method simple and scalable while preserving the sharpness of the projected result,” Iwai stated [1]. The University of Osaka, a national research institution that traces its roots to Edo-era schools and was formally established as an imperial university in 1931, has produced multiple IEEE Fellows in engineering fields [5][6][7]. In tabletop tests, the setup eliminated clearly defined shadows when a hand was placed near the surface, and brightness dropped only slightly when a fingertip touched it [1]. On a non-flat mannequin head, shadows were faintly visible but greatly reduced. The new blur-compensation technique computed results in roughly 100 seconds, compared with 2,500 seconds for a conventional system evaluated in the same study [1]. Seventeen participants took part in user studies. When the system projected text next to physically printed text, subjects were significantly slower and less accurate at identifying the projected content, suggesting the projections blended more convincingly with real objects [1]. Iwai noted that each projector currently requires individual calibration, and the team plans to explore distributed rendering in which small edge computers control individual projectors [1].